Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Scientific Reports volume 12, Article number: 14488 (2022)
Metrics details
Crop yield and its prediction are crucial in agricultural production planning. This study investigates and predicts arabica coffee yield in order to match the market demand, using artificial neural networks (ANN) and multiple linear regression (MLR). Data of six variables, including areas, productivity zones, rainfalls, relative humidity, and minimum and maximum temperature, were collected for the recent 180 months between 2004 and 2018. The predicted yield of the cherry coffee crop continuously increases each year. From the dataset, it was found that the prediction accuracy of the R2 and RMSE from ANN was 0.9524 and 0.0784 tons, respectively. The ANN model showed potential in determining the cherry coffee yields.
Coffee is one of the most popular global beverages, and Thailand is among the world’s top 25 coffee exporters, with over 30,000 tons per year1. It can generate enormous earnings for the country’s economy and farmers. It was found that high-standard coffee can get an asking price a few times more than standard products2,3. However, coffee products have been encountering several problems, including unqualified beans, climate change and environmental effects, a decrease in the nutrition level in soils, pests and insects, a lack of farming knowledge management, as well as the unavailability of raw materials4,5,6,7. While the main focus should be on the quality of the products to satisfy customer requirements, the lack of management might be the leading cause of these problems and might affect the efficiency of the accuracy rate. This could also subsequently lead to the failure of marketing prices in the future8,9. At present, the world’s coffee consumption has grown by an average of over 2% a year in the past decade10. The Thai coffee industry has shown similar growth between 2016 to 2020. Thailand’s average demand for coffee beans is almost 79,000 tons per year11. Accurate prediction of coffee will improve farmers’ or entrepreneurs’ ability to plan their annual coffee production and reduce the risk of inconsistent demand, supply, and lack of trade opportunities.
Arabica coffee (Coffea arabica L.), one of the main coffee species, has been considered a suitable plant to promote as an opium substitute to stop shifting cultivation12,13. However, Arabica is also one of the most popular species of coffee that customers look for because it features a particular characteristic flavor that differs from other types of coffee14. For this reason, farmers and entrepreneurs are likely to concentrate on Arabica coffee to meet the demand from customers. Hence, it is necessary to predict each year’s yields in order to manage the insufficient supply and the overwhelming demands of customers so as to reduce the loss of trade opportunities. Forecasting applications have recently been used to predict demand, supply management, and resources to reduce mistakes in terms of overproduction and operating costs15,16. Prediction in agriculture is mainly made in terms of rainfall estimates, area, production or productivity, and price17,18. Thus, these factors should be used to predict the productivity and yield of agricultural products, possibly with a machine learning model, as shown in Fig. 1.
Machine learning predictive models.
Machine learning (ML) is carried out to learn and make decisions without setting rules and conditions19 because it can learn through reading human-provided information and processing data. This data information can be applied to make the desired decisions20. The decision-making of ML has various formats, for example, predicting and classifying a new database21, identifying and clustering with different data22,23, detecting incorrect data information24, and introducing new data that are expected to interest humans by learning from previous interests25,26. In our problem, the multilayer perceptron was adopted for prediction. Neurons in the network are generated in a layer. Then, the output of one layer supports the input into the next layer and conceivably others. The model parameter of this study can be optimized by using the feed-forward backpropagation training algorithm. The ML model has great potential for application to coffee production prediction. Therefore, this research uses artificial neural networks (ANN) in coffee cherry prediction because of its ability to learn the patterns of complex systems. Moreover, it can predict future productivity from the current database more effectively than the traditional statistical method.
ML is critical to studying programming and computer design, such as robotic systems, visual systems, or computer vision27,28,29. A computer algorithm generates the machine learning program, and this program is then able to calculate and evaluate the condition, and the software will generate the decision-making. Then, the data is transferred to the dataset in order to generate the model. After that, the completed program is applied to determine work usability30. ML can be categorized into two groups; (i) supervised learning, which comprises regression and classification groups, and (ii) unsupervised learning, which comprises clustering and non-clustering groups31. ML uses a variety of algorithms, for example, linear regression, logistics regression, support vector machines, decision trees, random forest, boosting, convolutional neural network, gradient-boosted trees, and ANN32,33,34. In this work, we are interested in ANN.
The ANN model is now considered a vital data-modeling tool35. It is a calculation model that activates the functional structures of biological neural networks. A neuron is one part of the process running unit that acquires inputs and consequently operates these inputs to achieve outputs. The construction of the ANN includes many architectures, consisting of single-layer perceptrons, multilayer perceptrons, recurrent ANN, and self-organization mapping. Each of these is appropriate for different problems36.
The significant benefit of the ANN technique is the capability of modeling multiple outputs concurrently37. Improvement of the ANN model requires data generation for training or testing, selecting the optimal configuration, and validating the model on an independent data set38. The optimal configuration was found to be 1 and 2 hidden layers, 2–10 neutrons, and 1000–10,000 learning runs39. ANN is connected via a structure of single-layer perceptrons and consists of input, output, and hidden layers, which reconstruct the input through the output layer. ANN analysis was used to analyze the input and output data using the MATLAB Neural Network Toolbox. It includes input, hidden, and output layers, which are the central elements of each ANN structure40. The input layer will obtain the input data and normalize from datasets through neurons. Then, all of the data will be transmitted to the hidden layer41. Every neuron of the subsequent layer(s) will calculate the output data, transmitting a linear combination of all the neurons in the previous layers42.
ANN is a technique that uses data mining. A mathematical model is used to connect the information processing and computation and simulate the function of neural networks43. Neurons in computer systems include input and output, with neuron weights that can indicate the input and output weights. Then, the threshold value is the decision for transferring the output data through the neurons. After the collaboration work is finished, the results are shown in numerical calculations format36.
Previously, ANN and MLR have been employed in many various fields, such as irrigation groundwater quality parameters44, forecasting PM levels45, predicting SO2 concentration46, slope stability47, intrinsic solubility of generic drugs48, electrical conductivity49, mosquito abundances in urban areas50, estimating monthly rainfall distribution in Thailand51, prediction of human skin permeability52, etc. Presently, ANNs are used in medicine and health care53,54, medical image denoising55, face detection and recognition56, road and traffic signaling57, road sign detection and recognition58,59, and in modeling conventional and solar earth60. ANN and MLR are also currently used in agriculture, for instance, predicting organic potato yield37 and the yield in the hulless barley61.
It can also be used in the agricultural field to detect and classify plant diseases, crops, and pests, especially in coffee bean species and coffee cherry62. Ahmad et al.63 applied ANN to develop an automatic coffee sorting system based on image processing. Moreover, ANN was used to evaluate the degree of coffee roasting, predicting the sensory quality of a roasted coffee-flavored sterilized drink and recognizing the degree of coffee roasting64. ANN could be applied to predict the amount of agricultural production by using previous datasets in order to obtain correct information through a machine learning-based algorithm for annual coffee production65. In addition, ANN can predict not only agricultural products but also enter many types of data and extract knowledge outside of the primary dataset. Torkashvand et al.66 applied the ANN and MLR model to evaluate and improve the accuracy rate in predicting kiwifruit firmness. Dhyani67 used the ANN and MLR model to measure the rainfall to anticipate the crop in the following year. Furthermore, Etminan et al.68 utilized an ANN model to identify the best drought tolerance indices in agronomy and plant breeding by providing this valuable model in other biological contexts.
According to the literature, ANN is one of the tools that is suitable for predicting agricultural products because it can simulate the physical behavior of a complex system from the input data. The advantages of ANN are as follows; (i) it can give a more accurate prediction than general methods, (ii) the model can still be used even if there are errors in the actual data, (iii) any conflicting results can still be implemented got either are continuous, (iv) the analysis of results is fast after the learning process69. On the other hand, there are disadvantages of ANN; (i) a long training time can be necessary in order to find the best weight to learn, (ii) it is not easy to choose suitable parameters (the best topology), (iii) the learned function can be difficult to understand70.
MLR is used to study the collaboration of the relationships of two or more variables that affect another value. The regression includes various linear regressions and logistic regression algorithms that control the learning model to predict the input data into the program71. Its algorithms are more complicated but are also used in agricultural production, for example, ordinary least squares regression, multivariate adaptive regression splines, multi-linear regression, and locally estimated scatterplot smoothing72.
From the literature review, it was noticed that ANN and MLR are not yet utilized widely in agricultural production. Their applications are relatively novel in the study of agricultural production. To the authors’ knowledge, only a few works have deployed ANN and MLR to predict agricultural crop yield. So far, no work has been reported on applying machine learning to coffee cultivation. This study investigates and analyzes the prediction of cherry coffee and quantity control during plantation before processing the roasted coffee beans in Chiang Rai in northern Thailand, as shown in Fig. 2.
Coffee plantation area in Chiang Rai, Thailand. Figure has been generated by Mr. Yotsaphat Kittichotsatsawat, the first author of this manuscript using the Qgis program. Version 3.8.1. URL. https://www.qgis.org/en/site/index.html.
In this study, we provided the predicted yield of arabica coffee for 15 years from 2004 to 2018. Coffee is seasonally produced and available 6 months a year. Off-season, there is no yield. The dataset used in the study consists of factors related to crop yield, including the area, productivity zone, monthly rainfall, monthly RH, and monthly temperature. The data source is from the northern part of Thailand, which has been collected for 180 months. A summary of the statistical characteristics of the dataset is presented in Table 1. It was shown that the productivity zone had a positive skewness distribution of 0.413, and the temperature had a negative skewness distribution of − 0.467. The other variables were slightly skewed. The following subsection shows the results of the modeling. Productivity, also known as crop yield, is the variable to be predicted using the MLR. To apply MLR modeling, the correlation between pairs of input variables was examined to decide which variables to include in the model. The results are shown in Table 2.
In the MLR model, it is common to include the input variables with a high linear correlation with the response variable but to exclude ones that are highly correlated among themselves because of the multicollinearity problem. Table 2 shows that the area and productivity zone variables are highly positively correlated, while the variables Tmax and area and Tmax and RH are moderately positively and negatively correlated, respectively. These results might lead to a multicollinearity problem; however, in this study, we included all variables in the MLR, as the dataset is one hundred eighty observations and six variables. The results are shown in “Prediction results”.
The MLR, in which the equation was derived from the full model of six variables, is used to predict crop yield. The MLR equation, R2, and the Durbin‒Watson test are shown in Table 3. From Table 3, the model that includes six input datasets yields a high value of R2, which indicates that the coffee features substantially explain the change in the crop yield (92.35%). Meanwhile, the Durbin‒Watson test is 1.56, which reveals a suitable autocorrelation for the predictive errors. After data analysis, it was found that the R2 is 0.9235, RMSE is 0.0784, and MSE is 0.0061 ton, which is suitable for the model (see Fig. 3).
The R2 and RMSE (ton) of the scatter plot on the MLR model between the actual and predicted crop yield.
From Fig. 3, when the MLR model tested the input data and crop yield, it can be seen that the linear relationship was preserved between the variable input datasets and the crop yield. Thus, farmers and entrepreneurs could use this data set to predict cherry coffee productivity in the next period.
The ANN analysis was performed based on the previous 180 months’ input dataset. In all ANN analyses, the input and output had eight neurons and one layer. The number of neurons in the hidden layer was defined using trial and error to find the best ANN configuration for predicting the crop yield of cherry coffee. The ANN configuration determined to be optimal should minimize the MSE and RMSE and optimize R2. The ANN configuration from the network topology started from one and then consecutively increased.
From Table 4, the network properties will set the treatments of the (i) type of network, (ii) input and target data, (iii) training and adaptation learning function, (iv) performance function, (v) layer number, (vi) number of neurons, and (vii) transfer function. Feed-forward backpropagation was used for setting the network type. The number of hidden layers is fixed as two, numbered H1 and H2, and the number of neurons will set the number of processing elements. Then, these data will be trained in order to obtain the best ANN performance. After training the ANN model, the best of the various ANN configurations is with the second hidden layer and the eighth processing element. Meanwhile, the MSE and R2 are 1027.99 and 0.9524 tons, respectively, as shown in Figs. 4 and 5.
Mean squared error (MSE) of cherry coffee validation performance.
Prediction of the coffee cherry using ANN (MSE training and predicted values of cherry coffee by multilayer perceptron neural network versus data value, training and validation phases) (ton).
Figure 4 shows the MSE from the random data generated, which included the training, validation, and testing data. The learning rate and momentum were constant at 1000, 0.5, and 0.1, respectively. The results show that the learning cycles were 49 epochs, and the best validation performance was found at 43 epochs with the MSE at 1027.99. Table 5 shows the outcome of the coefficient (R2) of the training and testing data phases. The neural network confirms the precise tools with perfect prediction accuracy in the responses of genotypes for coffee crop yield based on the measured indices. Thus, the ANN model was demonstrated to provide a perfect agreement between the exact and approximate coffee productivity. However, the prediction potential of productivity by ANN (0.9524) appeared to show higher accuracy than MLR (0.9235). MLR and ANN were compared through machine learning techniques using historical areas, productivity zones, rainfalls, relative humidity, and minimum and maximum temperature for the recent 180 months.
According to the study, MLR and ANN have focused on practical interpolation methods. RMSE and MSE cross-validation were accomplished to assess the efficiency of each month (168 months). It was also estimated through the average of RMSE and MSE. However, the cross-validation of MLR showed that the average values of RMSE and MSE are 0.0618 and 0.0038, respectively. Nonetheless, in comparison to the full model of coffee prediction for 180 months, RMSE was 0.0784, and MSE was 0.0061 tons, respectively. The results showed that both of them were not significantly different.
The crop yield of cherry coffee varies because of several factors such as the cultivated area, amount of rainfall, temperature, and RH. These conditions affect the cherry coffee yield from each month’s crop. Prediction is necessary to forecast the productivity accurately and to satisfy customer demand. Previously, farmers and entrepreneurs lack knowledge regarding the future cherry coffee yield, as well as which factors can change the cherry coffee crop yield each month. For this reason, production planning to match the supply to the demand is rather difficult. However, if farmers and entrepreneurs have sufficient data and can utilize it using machine learning models, they may realize more accurate estimate of the crop yield.
In this work, MLR and ANN were used to analyze and predict the coffee crop yield using data from 2004 to 2018, in which input data for 132 months was used for training, 24 months for validation, and 24 months for testing. Calculation of the cherry coffee crop yield required MATLAB programming. The MLR equation was derived from the full model consisting of six variables. The data shows that the area (X1) ranged between 42,216 and 20,158 acres with an average of 34,003 acres, playing an essential role in cherry coffee production. The available area is a crucial component for coffee tree planting because it determines the farmer’s cherry coffee productivity. The productivity zone (X2) shows data ranging between 37,710 and 14,513 acres, with an average of 23,921 acres. This is also essential because it indicates how much the farmer can earn from the cherry coffee yield73. The rainfall data (X3) is from 0 to 509.8, but the average rainfall is 1830 mm per month. The amount of RH (X4) is between 55.0 and 82.0%RH, and the mean is 76.0%RH which is suitable for coffee cultivation74. Moreover, the temperature (X5 and X6) is a factor that could affect coffee production and is shown in the data to be between 5.10 and 40.0 °C with an average of 25.5 °C75.
The correlation was checked between independent variables. As in Table 2, there are some high correlations, such as those between the minimum temperature and productivity, the area and productivity zone, and the productivity zone and productivity, with values of 0.9175, 0.8119, and 0.7775, respectively. For this reason, it was necessary to put all the variables into the model or may represent them with VIF < 10 values (variance inflation factor). The assumption of MLR was adjusted and estimated from (widehat{Y})‘ = 0.2850 + 0.0402X1 + 0.3128X2 − 0.0486X3 − 0.4040X4 − 0.6144X5 + 1.1065X6 with 0.9235 for the standard error of estimation. The R2 and Durbin‒Watson values confirmed the appropriateness of the MLR equation, as shown in Table 3. R2 and Durbin‒Watson values were 0.9235 and 1.56, respectively which are suitable for this MLR model. Meanwhile, R2 in Table 3 also indicates that the coffee variable substantially explains the changes in the crop yield, while the Durbin‒Watson test reveals a low autocorrelation for the predictive errors.
The importance of each input variable on the yield was also analyzed based on the MLR mathematical expression. The equation coefficients showed that Tmin monthly variable (X6) was the most important (1.1065). The second-most important factor was Tmax monthly (X5) with a coefficient of − 0.6144. The relative humidity (X4), the productivity zone (X2), and monthly total rainfall (X3) were inferior, with values of − 0.4040, 0.3128, and − 0.0486, respectively. Lastly, the area (X1) was shown to be the least essential variable (0.0402).
Similarly, for the ANN model, the relative importance of each input variable was evaluated based on one-way partial dependence plots (PDPs), as shown in Fig. 6. It can be seen that the monthly Tmax variable (X5) was the most important, in which the PDP value varied from 0.45 to 0.05. Similarly, the second most influential variable was monthly total rainfall (X3), with PDP values of 0.25 to 0.05. The third variable was relative humidity (X4) of 0.138 to 0.103. Additionally, the last three variable groups were productivity zone (X2), area (X1), and Tmin monthly (X6), showing the PDP values varied from 0.240 to 0.07, 0.158 to 0.118, and 0.160 to 0.110, respectively76,77. In Fig. 6e, when the Tmax monthly variable increased, cheery coffee’s productivity was lowered, especially when it was higher than 29 °C. In other words, if the temperature was less than 29 °C, the productivity of cheery coffee was improved. From Fig. 6b, if we have more productivity zones, productivity will increase accordingly. From Fig. 6d, the productivity would be high if the cherry coffee experienced high relative humidity. On the contrary, if the amount of rain were too large, it would negatively affect productivity. Based on the PDPs in Fig. 6c, the cherry coffee should rain at less than 100 mm per month, leading to a high crop yield.
One-way PDPs of input variables on prediction of crop yields.
Likewise, the ANN method can obtain the input dataset from the available data. Then, the data was imported from the MATLAB workspace, and a variable was selected. A network or data was created to random the hidden layer until the performances of the various ANN configurations were known. The best performance is shown in Table 5, with values of 0.9524 and 0.0784 ton for R2 and RMSE, respectively. The results of both models indicate that the cherry coffee crop yield can be accurately predicted if the farmers or entrepreneurs are able to utilize these data.
In comparing MLR and ANN performances with other agricultural-related predictions, the R2 of the MLR was better than the ANN in some cases due to its limited number of input datasets. This was so for Bayat et al. work78 on predicting tree survival and mortality in the Hyrcanian forest. The six input variables were used for 9 years of prediction, including the wind velocity and direction, diameter at breast height, basal area of larger trees, diameter, topographic wetness index, and temperature. The R2 of ANN and MLR were 0.6790 and 0.9200, respectively. Ustaoglu et al.79 used MLR and ANN for temperature forecasts through daily mean, maximum, and minimum temperature in time series. The R2 of ANN was 0.8590, lower than that of MLR (0.9025).
The R2 of ANN was generally found to be better than that of MLR when many datasets were used, for example, in predicting rainfall75,80, and maize yield through the temperature and precipitation data81, monthly pan evaporation and soil temperatures82,83, the relative humidity of climate in Turkey84, and crop prediction85,86, as shown in Table 6. For the cherry coffee crop yield by MLR and ANN in this work, the R2 were 0.9235 and 0.9524, respectively, while RMSE was 0.0784 ton, showing the suitability for crop yield prediction.
MLR and ANN were effective in forecasting the crop yield of cherry coffee. These methods can be utilized from upstream to downstream in coffee planning to respond to customer demand. Moreover, they can reduce the risk of insufficient production in the high season and improve future efficiency and effectiveness for industrial and agricultural products.
Concerning limitations, this research considered only six input datasets, which were the area, amount of rainfall, temperature, and RH that affected the efficiency of increased crop yield. More related data input such as farm location, irrigation water depth applied, precipitation, solar radiation, potential evapotranspiration, soil moisture, land cultivated, pH, nitrogen, phosphorus, potassium, air pressure, wind speed, and climate variables should also be considered. The crop yield prediction could be improved further87,88,89.
Once the crop yield and amount of cherry coffee are accurately predicted for future works, farmers or entrepreneurs can prepare their manufacturing resources for later stages of production planning accordingly. These may include (i) color sorters to classify the color shade of cherry coffee for higher added value and (ii) dryer machines to reduce the moisture content in the cherry coffee.
The ANN algorithm has been demonstrated to be effective in making predictions of production rates/yields from industries of coffee plantings. However, the dataset used may be imbalanced. It is essential to note that, in identifying the optimal granularity and refining the imbalanced dataset, combining granular computing and ML techniques is of great interest90,91. In general, the predictions of developed ML models are based on knowledge learning from the complex multi-dimensional relationships between features and target(s)92. When an imbalanced, small dataset is employed to train and develop an ML model, the developed model would have a poor ability to make predictions. An innovative approach of combining granular computing with ML or deep learning may be considered further, in which the dataset’s quality and efficiency could be improved via clustering examples in the best granularity, sampling, and refining into a more balanced example set, such as those deployed in services planning under social manufacturing context and bi-level model for risk classification of respiratory diseases90,91,92,93,94. Additionally, reinforcement learning techniques may be considered to predict agricultural production. The learning process happens along the way with six main components of (i) agent, (ii) action, (iii) environment, (iv) state, (v) policy, and (vi) reward. It was successful for order acceptance decision of mass-individualized printed circuit board manufacturing95. In this research, only six main input features were considered. Therefore, reinforcement learning with more complex datasets may be considered for cherry coffee and other agricultural industries in the near future.
This study focuses on predicting cherry coffee. The scope of this study covers four steps from plantation until harvesting (see Fig. 7).
Research framework for cherry coffee prediction.
The data from the Climate Department included rainfall, RH, and minimum and maximum temperature. The Agricultural Economics Office and Meteorological Department provided the area and productivity zone data. The input dataset consisted of the area, rainfall, temperature, and (RH), while the output consisted of the crop yield of cherry coffee. The input and output data were normalized within the range 0‒1 using
where N is the normalized data; X is the measured value: Xmin and Xmax are the minimum and maximum values. The ANN performance must be estimated by means of a validation dataset. The selected ANN was tested, and its performance was compared using the coefficient of determination (R2), the mean squared error (MSE), and the percentage error.
MLR is one of the supervised learning groups that can generate the model for constructing a relationship between two or more interpretive variables (independent) and response variables (dependent) via appropriate linear equations into the observed data. The model for MLR is:
where Yi is the dependent variable; b0 is a constant (intercept); Xi,k is an independent variable; bk is the vector of regression coefficients (slope); and ei is random measured errors.
The variables data were randomly split into three groups to prevent overfitting in the training step because a high number of neurons existed in the hidden layer. Training, validation, and testing of models were based on 70%, 15%, and 15% of the data, respectively68. In the meantime, during neural network training, the algorithm provided the minimum or maximum performance via the shortest path to yield the network’s size. At the same time, backpropagation was performed through the training set in order to update the rule to seek and find the minimum mean square error over the training set.
The best point of generalization form validation was obtained from a training stop when the error started to increase until the final increase point. Next, the networks were trained with the testing set, which was also compared with the output dataset with the desired output, and the model weight was fixed96. After the training epoch, the mean square error of the network running of the validation set was calculated. Then, the neurons will apply a specific linear function in the hidden layer and a specific linear function in a hidden layer to collect the linear combination and bias. Finally, the output data will give the predicted model.
The validated model was evaluated based on the observed and predicted coffee crop yield difference. The 180 months of data on coffee production were randomly evaluated through a k-fold cross-validation technique but evenly partitioned into 168 months of the dataset. The dataset will be further randomly evaluated through three rounds of MLR and ANN (ten processing elements and two hidden layers) training. Each dataset was individually used as validation, while the enduring ones were treated as training datasets. From the k-fold cross-validation, RMSE and MSE were evaluated to determine the models’ performance:
where n is the sample size of the testing dataset, while E (xi) and M(xi) are interpolated and observed values, respectively.
In this model, the crop yield was considered to assess indices through the input and output data. A partial dependence plot (PDP) was performed to evaluate each variable as an independent variable during the neural network process76,77. The relative importance of the parameters for each index was computed by dividing the value of its importance by the highest importance value. Consequently, the model with the largest R2 and the smallest RMSE was considered the best.
The ratios of the correlation between standard errors and mean values were calculated by evaluating the accuracy in terms of repeatability and similarity for yield-based with the maximal indices up to 1000 bootstrap samples similar to the coefficient of variation (CV). A low CV indicates a selection index that is more accurate65. All implementations were accomplished using the neural network toolbox in the MATLAB package. Therefore, in the present study, MATLAB version R2019b, with Windows 10 and a 64-bit machine, was used to run the MLR and ANN models.
The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.
Lewin, B., Giovannucci, D. & Varangis, P. Coffee markets: New paradigms in global supply and demand. World Bank Agric. Rural Dev. Discuss. Paper https://doi.org/10.2139/ssrn.996111 (2004).
Article Google Scholar
Panhuysen, S. & Pierrot, J. Coffee barometer 2014. Hivos, IUCN Nederland, Oxfam, Novib, Solidaridad, WWF. (2014).
Torok, A., Mizik, T., Jambor, A. J. I. J. O. E. & Issues, F. The competitiveness of global coffee trade. Int. J. Econ. Financ. 8, 1–6. https://doi.org/10.32479/ijefi.6692 (2018).
CAS Article Google Scholar
Degaga, J. & Alamerie, K. J. E. Supply and performance of coffee markets in Gololcha district of Oromia region, Ethiopia. Ekonomika Poljoprivrede. 67, 797–816. https://doi.org/10.5937/ekopolj2003797d (2020).
Article Google Scholar
Genanaw, T. & Lamenew, W. Indigenous knowledge management framework for coffee production in Ethiopia. Ethiop. e-J. Res. Innov. Foresight. 6, 53–61 (2019).
Google Scholar
Duarte, A., Uribe, J. C., Sarache, W. & Calderón, A. J. E. Economic, environmental, and social assessment of bioethanol production using multiple coffee crop residues. Energy 216, 119170. https://doi.org/10.1016/j.energy.2020.119170 (2021).
CAS Article Google Scholar
Effendi, D. & Rismaya, M. Design and development of coffee production information system to support coffee production productivity in farmers group. IOP Conf. Ser. Mater. Sci. Eng. 879, 012058. https://doi.org/10.1088/1757-899x/879/1/012058 (2020).
Article Google Scholar
Baloi, D. & Price, A. D. J. I. J. O. P. M. Modelling global risk factors affecting construction cost performance. Int. J. Proj. Manag. 21, 261–269. https://doi.org/10.1016/s0263-7863(02)00017-0 (2003).
Article Google Scholar
Kittichotsatsawat, Y., Jangkrajarng, V. & Tippayawong, K. Y. J. S. Enhancing coffee supply chain towards sustainable growth with big data and modern agricultural technologies. Sustainability. 13, 4593. https://doi.org/10.3390/su13084593 (2021).
Article Google Scholar
International Coffee Organization. Developing a sustainable coffee economy (2020). http://www.ico.org/sustaindev_e.asp. Accessed 30 March 2022.
Chuqian, W. A study on the situation and development of the coffee industry in Thailand (2018).
Trébuil, G., Ekasingh, B. & Ekasingh, M. Agricultural commercialisation, diversification, and conservation of renewable resources in northern Thailand highlands. Moussons. 9–10, 131–155. https://doi.org/10.4000/moussons.2005 (2006).
Article Google Scholar
Sutthi, C. Highland agriculture: From better to worse. Hill Tribes Today. 107–142. (1989).
Romani, S., Cevoli, C., Fabbri, A., Alessandrini, L. & Dalla Rosa, M. Evaluation of coffee roasting degree by using electronic nose and artificial neural network for off-line quality control. J. Food Sci. 77, C960–C965. https://doi.org/10.1016/j.foodres.2020.109667 (2012).
CAS Article PubMed Google Scholar
Moon, M. A. Demand and supply integration: The key to world-class demand forecasting. Walter de Gruyter https://doi.org/10.1515/9781501506024 (2018).
Article Google Scholar
Doucoure, B., Agbossou, K. & Cardenas, A. Time series prediction using artificial wavelet neural network and multi-resolution analysis: Application to wind speed data. Renew. Energy. 92, 202–211. https://doi.org/10.1016/j.renene.2016.02.003 (2016).
Article Google Scholar
Ashokkumar, K., Chowdary, D. D. & Sree, C. D. Data analysis and prediction on cloud computing for enhancing productivity in agriculture. IOP Conf. Ser. Mater. Sci. Eng. 590, 012014. https://doi.org/10.1088/1757-899x/590/1/012014 (2019).
Article Google Scholar
Rajeswari, S. & Suthendran, K. Developing an agricultural product price prediction model using HADT algorithm. Int. J. Eng. Adv. Technol. 9, 569–575. https://doi.org/10.35940/ijeat.a1126.1291s419 (2019).
Article Google Scholar
Janiesch, C., Zschech, P. & Heinrich, K. Machine learning and deep learning. Electron. Mark. 3, 685–695. https://doi.org/10.1007/s12525-021-00475-2 (2021).
Article Google Scholar
Tripathy, A., Agrawal, A. & Rath, S. K. Classification of sentiment reviews using n-gram machine learning approach. Expert Syst. Appl. 57, 117–126. https://doi.org/10.1016/j.eswa.2016.03.028 (2016).
Article Google Scholar
Sun, S., Cao, Z., Zhu, H. & Zhao, J. A survey of optimization methods from a machine learning perspective. IEEE Trans. Cybern. 8, 3668–3681. https://doi.org/10.1109/tcyb.2019.2950779 (2019).
Article Google Scholar
Valletta, J. J., Torney, C., Kings, M., Thornton, A. & Madden, J. Applications of machine learning in animal behaviour studies. Anim. Behav. 124, 203–220. https://doi.org/10.1016/j.anbehav.2016.12.005 (2017).
Article Google Scholar
Dogan, A. & Birant, D. Machine learning and data mining in manufacturing. Expert Syst. Appl. 166, 114060. https://doi.org/10.1016/j.eswa.2020.114060 (2021).
Article Google Scholar
Wiens, J. & Shenoy, E. S. Machine learning for healthcare: On the verge of a major shift in healthcare epidemiology. Clin. Infect. Dis. 66, 149–153. https://doi.org/10.1093/cid/cix731 (2018).
Article PubMed Google Scholar
Howard, J. Artificial intelligence: Implications for the future of work. Am. J. Ind. Med. 11, 917–926. https://doi.org/10.1002/ajim.23037 (2019).
Article Google Scholar
L’heureux, A., Grolinger, K., Elyamany, H. F. & Capretz, M. A. Machine learning with big data: Challenges and approaches. IEEE Access. 5, 7776–7797. https://doi.org/10.1109/access.2017.2696365 (2017).
Article Google Scholar
Kim, D. et al. Review of machine learning methods in soft robotics. PLoS ONE 16, e0246102. https://doi.org/10.1371/journal.pone.0246102 (2021).
CAS Article PubMed PubMed Central Google Scholar
Abiodun, O. I. et al. State-of-the-art in artificial neural network applications: A survey. Heliyon. 11, e00938. https://doi.org/10.1016/j.heliyon.2018.e00938 (2018).
Article Google Scholar
Lindsay, G. W. Attention in psychology, neuroscience, and machine learning. Front. Comput. Neurosci. 14, 29. https://doi.org/10.3389/fncom.2020.00029 (2020).
Article PubMed PubMed Central Google Scholar
Dalal, S. R. et al. In model-based testing in practice. In Proceedings of the 21st International Conference on Software Engineering—ICSE ’99. 285–294. https://doi.org/10.1145/302405.302640 (1999).
Brownlee, J. Supervised and unsupervised machine learning algorithms. Mach. Learn. Mastery. 3 (2016). https://machinelearningmastery.com/supervised-and-unsupervised-machine-learning-algorithms/. Accessed 30 March 2022.
Çınar, Z. M. et al. Machine learning in predictive maintenance towards sustainable smart manufacturing in industry 4.0. Sustainability. 19, 8211. https://doi.org/10.3390/su12198211 (2020).
Article Google Scholar
Onsree, T. & Tippayawong, N. Machine learning application to predict yields of solid products from biomass torrefaction. Renew. Energy. 167, 425–432. https://doi.org/10.1016/j.renene.2020.11.099 (2021).
CAS Article Google Scholar
Onsree, T., Tippayawong, N., Phithakkitnukoon, S. & Lauterbach, J. Interpretable machine-learning model with a collaborative game approach to predict yields and higher heating value of torrefied biomass. Energy 249, 123676. https://doi.org/10.1016/j.energy.2022.123676 (2022).
Article Google Scholar
Agatonovic-Kustrin, S. & Beresford, R. Basic concepts of artificial neural network (ANN) modeling and its application in pharmaceutical research. J. Pharm. Biomed. Anal. 5, 717–727. https://doi.org/10.1016/s0731-7085(99)00272-1 (2000).
Article Google Scholar
Singh, Y. & Chauhan, A. S. Neural networks in data mining. J. Theor. Appl. Inf. Technol. 1, 1–6 (2009).
CAS Google Scholar
Abrougui, K. et al. Prediction of organic potato yield using tillage systems and soil properties by artificial neural network (ANN) and multiple linear regressions (MLR). Soil Tillage Res. 190, 202–208. https://doi.org/10.1016/j.still.2019.01.011 (2019).
Article Google Scholar
May, R. J., Dandy, G. C., Maier, H. R. & Nixon, J. B. Application of partial mutual information variable selection to ANN forecasting of water quality in water distribution systems. Environ. Model. Softw. 10–11, 1289–1299. https://doi.org/10.1016/j.envsoft.2008.03.008 (2008).
Article Google Scholar
Anderson, D. & McNeill, G. Artificial neural networks technology. Kaman Sci. Corp. 6, 1–83 (1992).
Google Scholar
Zain, A. M., Haron, H. & Sharif, S. Prediction of surface roughness in the end milling machining using artificial neural network. Expert Syst. Appl. 37, 1755–1768. https://doi.org/10.1016/j.eswa.2009.07.033 (2010).
Article Google Scholar
Togun, N. K. & Baysec, S. Prediction of torque and specific fuel consumption of a gasoline engine by using artificial neural networks. Appl. Energy. 87, 349–355. https://doi.org/10.1016/j.apenergy.2009.08.016 (2010).
CAS Article Google Scholar
Yosinski, J., Clune, J., Nguyen, A., Fuchs, T. & Lipson, H. Understanding neural networks through deep visualization. Explain AI. Interpret. Explain. Vis. Deep Learn. https://doi.org/10.1007/978-3-030-28954-6_4 (2015).
Article Google Scholar
Krenker, A., Bešter, J. & Kos, A. Introduction to the artificial neural networks. Artif. Neural Netw. Methodol. Adv. Biomed. Appl https://doi.org/10.5772/15751 (2011).
Article Google Scholar
Kouadri, S., Pande, C. B., Panneerselvam, B., Moharir, K. N. & Elbeltagi, A. Prediction of irrigation groundwater quality parameters using ANN, LSTM, and MLR models. Environ. Sci. Pollut. Res. 29, 21067–21091. https://doi.org/10.1007/s11356-021-17084-3 (2022).
CAS Article Google Scholar
Ceylan, Z. & Bulkan, S. Forecasting PM10 levels using ANN and MLR: A case study for Sakarya City. Glob. Nest J. 20, 281–290. https://doi.org/10.30955/gnj.002522 (2018).
CAS Article Google Scholar
Shams, S. R., Jahani, A., Kalantary, S., Moeinaddini, M. & Khorasani, N. The evaluation on artificial neural networks (ANN) and multiple linear regressions (MLR) models for predicting SO2 concentration. Urban Clim. 37, 100837. https://doi.org/10.1016/j.uclim.2021.100837 (2021).
Article Google Scholar
Chakraborty, A. & Goswami, D. Prediction of slope stability using multiple linear regression (MLR) and artificial neural network (ANN). Arab. J. Geosci. 10, 1–11. https://doi.org/10.1007/s12517-017-3167-x (2017).
Article Google Scholar
Louis, B., Agrawal, V. K. & Khadikar, P. V. Prediction of intrinsic solubility of generic drugs using MLR, ANN and SVM analyses. Eur. J. Med. Chem. 45, 4018–4025. https://doi.org/10.1016/j.ejmech.2010.05.059 (2010).
CAS Article PubMed Google Scholar
Ekemen Keskin, T., Özler, E., Şander, E., Düğenci, M. & Ahmed, M. Y. Prediction of electrical conductivity using ANN and MLR: A case study from Turkey. Acta Geophys. 68, 811–820. https://doi.org/10.1007/s11600-020-00424-1 (2020).
ADS Article Google Scholar
Lee, K. Y., Chung, N. & Hwang, S. Application of an artificial neural network (ANN) model for predicting mosquito abundances in urban areas. Ecol. Inform. 36, 172–180. https://doi.org/10.1016/j.ecoinf.2015.08.011 (2016).
Article Google Scholar
Chutsagulprom, N., Chaisee, K., Wongsaijai, B., Inkeaw, P. & Oonariya, C. Spatial interpolation methods for estimating monthly rainfall distribution in Thailand. Theor. Appl. Climatol. 148, 1–12. https://doi.org/10.21203/rs.3.rs-568778/v1 (2022).
Article Google Scholar
Chen, L. J., Lian, G. P. & Han, L. J. Prediction of human skin permeability using artificial neural network (ANN) modeling. Acta Pharmacol. Sin. 28, 591–600. https://doi.org/10.1111/j.1745-7254.2007.00528.x (2007).
CAS Article PubMed Google Scholar
Soroushmehr, S. R. & Najarian, K. Transforming big data into computational models for personalized medicine and health care. Dialogues Clin. Neurosci. https://doi.org/10.31887/dcns.2016.18.3/ssoroushmehr (2022).
Article Google Scholar
Cervantes-Sanchez, F., Cruz-Aceves, I., Hernandez-Aguirre, A., Hernandez-Gonzalez, M. A. & Solorio-Meza, S. E. Automatic segmentation of coronary arteries in X-ray angiograms using multiscale analysis and artificial neural networks. Appl. Sci. 24, 5507. https://doi.org/10.3390/app9245507 (2019).
Article Google Scholar
Priya, D. K., Sam, B. B., Lavanya, S. & Sajin, A. P. A survey on medical image denoising using optimisation technique and classification. Int. Conf. Inf. Commun. Embed. Syst. https://doi.org/10.1109/icices.2017.8070729 (2017).
Article Google Scholar
Patil, J. V. & Bailke, P. Real time facial expression recognition using RealSense camera and ANN. Int. Conf. Invent. Comput. Technol. 2, 1–6. https://doi.org/10.1109/inventive.2016.7824820 (2016).
Article Google Scholar
Islam, K. T. & Raj, R. G. Real-time (vision-based) road sign recognition using an artificial neural network. Sensors. 17, 853. https://doi.org/10.3390/s17040853 (2017).
ADS Article PubMed Central Google Scholar
Lahmyed, R., Ansari, M. E. & Kerkaou, Z. J. S. C. Automatic road sign detection and recognition based on neural network. Soft Comput. 26, 1–22. https://doi.org/10.1007/s00500-021-06726-w (2022).
Article Google Scholar
Kryvinska, N., Poniszewska-Maranda, A. & Gregus, M. An approach towards service system building for road traffic signs detection and recognition. Proc. Comput. Sci. 141, 64–71. https://doi.org/10.1016/j.procs.2018.10.150 (2018).
Article Google Scholar
Chauhan, R., Dumka, P. & Mishra, D. R. Modelling conventional and solar earth still by using the LM algorithm-based artificial neural network. Int. J. Ambient Energy. 43, 1389–1396. https://doi.org/10.1080/01430750.2019.1707113 (2022).
CAS Article Google Scholar
Zaefizadeh, M., Jalili, A., Khayatnezhad, M., Gholamin, R. & Mokhtari, T. Comparison of multiple linear regressions (MLR) and artificial neural network (ANN) in predicting the yield using its components in the hulless barley. Adv. Environ. Biol. 5, 109–114 (2011).
Google Scholar
Anita, S. In classification Cherry’s Coffee using k-Nearest Neighbor (KNN) and Artificial Neural Network (ANN). In 2020 International Conference on Information Technology Systems and Innovation (ICITSI).117–122 (IEEE). https://doi.org/10.1109/icitsi50517.2020.9264925 (2020).
Ahmad, A. S. et al. A review on applications of ANN and SVM for building electrical energy consumption forecasting. Renew. Sustain. Energy Rev. 33, 102–109. https://doi.org/10.1016/j.rser.2014.01.069 (2014).
ADS Article Google Scholar
Leme, D. S., da Silva, S. A., Barbosa, B. H. G., Borém, F. M. & Pereira, R. G. F. A. Recognition of coffee roasting degree using a computer vision system. Comput. Electron. Agric. 156, 312–317. https://doi.org/10.1016/j.compag.2018.11.029 (2019).
Article Google Scholar
Van Klompenburg, T., Kassahun, A. & Catal, C. Crop yield prediction using machine learning: A systematic literature review. Comput. Electron. Agric. 177, 105709. https://doi.org/10.1016/j.compag.2020.105709 (2020).
Article Google Scholar
Torkashvand, A. M., Ahmadi, A. & Nikravesh, N. L. Prediction of kiwifruit firmness using fruit mineral nutrient concentration by artificial neural network (ANN) and multiple linear regressions (MLR). J. Integr. Agric. 7, 1634–1644. https://doi.org/10.1016/s2095-3119(16)61546-0 (2017).
CAS Article Google Scholar
Dhyani, S. Predicting Rainfall for Agriculture in India Using Regression (Dublin Business School, 2020). https://esource.dbs.ie/handle/10788/4230.
Etminan, A. et al. Determining the best drought tolerance indices using artificial neural network (ANN): Insight into application of intelligent agriculture in agronomy and plant breeding. Cereal Res. Commun. 47, 170–181. https://doi.org/10.1556/0806.46.2018.057 (2019). Accessed 30 March 2022.
Article Google Scholar
Afan, H. A. et al. ANN based sediment prediction model utilizing different input scenarios. Water Resour. Manag. 4, 1231–1245. https://doi.org/10.1007/s11269-014-0870-1 (2015).
Article Google Scholar
Bouktif, S., Fiaz, A., Ouni, A. & Serhani, M. A. Optimal deep learning lstm model for electric load forecasting using feature selection and genetic algorithm: Comparison with machine learning approaches. Energies 7, 1636. https://doi.org/10.3390/en11071636 (2018).
Article Google Scholar
Landwehr, N., Hall, M. & Frank, E. Logistic model trees. Mach. Learn. Mastery. 1, 161–205. https://doi.org/10.1007/s10994-005-0466-3 (2005).
Article MATH Google Scholar
Wilkinson, L. Statistical methods in psychology journals: Guidelines and explanations. Am. Psychol. 8, 594. https://doi.org/10.1037/0003-066x.54.8.594 (1999).
Article Google Scholar
Kath, J., Byrareddy, V. M., Mushtaq, S., Craparo, A. & Porcel, M. Temperature and rainfall impacts on robusta coffee bean characteristics. Clim. Risk Manag. 32, 100281. https://doi.org/10.1016/j.crm.2021.100281 (2021).
Article Google Scholar
Jayakumar, M., Rajavel, M. & Surendran, U. Climate-based statistical regression models for crop yield forecasting of coffee in humid tropical Kerala, India. Int. J. Biometeorol. 12, 1943–1952. https://doi.org/10.1007/s00484-016-1181-4 (2016).
Article Google Scholar
Ilaboya, I. R. & Igbinedion, O. E. Performance of multiple linear regression (MLR) and artificial neural network (ANN) for the prediction of monthly maximum rainfall in Benin City, Nigeria. Int. J. Eng. Sci. Appl. 1, 21–37 (2019).
Google Scholar
Katongtung, T., Onsree, T. & Tippayawong, N. Machine learning prediction of biocrude yields and higher heating values from hydrothermal liquefaction of wet biomass and wastes. Bioresour. Technol. 344, 126278. https://doi.org/10.1016/j.biortech.2021.126278 (2022).
CAS Article PubMed Google Scholar
Wongchai, W., Onsree, T., Sukkam, N., Promwungkwa, A. & Tippayawong, N. Machine learning models for estimating above ground biomass of fast growing trees. Expert Syst. Appl. 199, 117186. https://doi.org/10.1016/j.eswa.2022.117186 (2022).
Article Google Scholar
Bayat, M., Ghorbanpour, M., Zare, R., Jaafari, A. & Pham, B. T. Application of artificial neural networks for predicting tree survival and mortality in the Hyrcanian forest of Iran. Comput. Electron. Agric. 164, 104929. https://doi.org/10.1016/j.compag.2019.104929 (2019).
Article Google Scholar
Ustaoglu, B., Cigizoglu, H. & Karaca, M. Forecast of daily mean, maximum and minimum temperature time series by three artificial neural network methods. Meteorol. Appl. 15, 431–445. https://doi.org/10.1002/met.83 (2008).
ADS Article Google Scholar
El-Shafie, A., Mukhlisin, M., Najah, A. A. & Taha, M. R. Performance of artificial neural network and regression techniques for rainfall-runoff prediction. Int. J. Phys. Sci. 6, 1997–2003. https://doi.org/10.54302/mausam.v62i1.4711 (2011).
Article Google Scholar
Matsumura, K., Gaitan, C. F., Sugimoto, K., Cannon, A. J. & Hsieh, W. W. Maize yield forecasting by linear regression and artificial neural networks in Jilin, China. J. Agric. Sci. 153, 399–410. https://doi.org/10.1017/s0021859614000392 (2015).
Article Google Scholar
Patle, G., Chettri, M. & Jhajharia, D. Monthly pan evaporation modelling using multiple linear regression and artificial neural network techniques. Water Supply 20, 800–808. https://doi.org/10.2166/ws.2019.189 (2020).
Article Google Scholar
Kisi, O., Tombul, M. & Kermani, M. Z. Modeling soil temperatures at different depths by using three different neural computing techniques. Theor. Appl. Climatol. 121, 377–387. https://doi.org/10.1007/s00704-014-1232-x (2015).
ADS Article Google Scholar
Yasar, A., Simsek, E., Bilgili, M., Yucel, A. & Ilhan, I. Estimation of relative humidity based on artificial neural network approach in the Aegean Region of Turkey. Meteorol. Atmos. Phys. 115, 81–87. https://doi.org/10.1007/s00703-011-0168-2 (2012).
ADS Article Google Scholar
Li, Y. et al. Toward building a transparent statistical model for improving crop yield prediction: Modeling rainfed corn in the US. Field Crops Res. 234, 55–65. https://doi.org/10.1016/j.fcr.2019.02.005 (2019).
Article Google Scholar
Han, X. et al. Crop evapotranspiration prediction by considering dynamic change of crop coefficient and the precipitation effect in back-propagation neural network model. J. Hydrol. 596, 126104. https://doi.org/10.1016/j.jhydrol.2021.126104 (2021).
Article Google Scholar
Adisa, O. M. et al. Application of artificial neural network for predicting maize production in South Africa. Sustainability. 4, 1145. https://doi.org/10.3390/su11041145 (2019).
Article Google Scholar
Gonzalez-Sanchez, A., Frausto-Solis, J. & Ojeda-Bustamante, W. Attribute selection impact on linear and nonlinear regression models for crop yield prediction. Sci. World J. 2014, 1–10. https://doi.org/10.1155/2014/509429 (2014).
Article Google Scholar
Balakrishnan, N. & Muthukumarasamy, G. Crop production-ensemble machine learning model for prediction. Int. J. Comput. Sci. Softw. Eng. 7, 148 (2016).
Google Scholar
Leng, J. et al. A loosely-coupled deep reinforcement learning approach for order acceptance decision of mass-individualized printed circuit board manufacturing in industry 4.0. J. Clean. Prod. 280, 124405. https://doi.org/10.1016/j.jclepro.2020.124405 (2021).
Article Google Scholar
Leng, J., Chen, Q., Mao, N. & Jiang, P. Combining granular computing technique with deep learning for service planning under social manufacturing contexts. Knowl. Based Syst. 143, 295–306. https://doi.org/10.1016/j.knosys.2017.07.023 (2018).
Article Google Scholar
Phromphithak, S., Onsree, T. & Tippayawong, N. Machine learning prediction of cellulose-rich materials from biomass pretreatment with ionic liquid solvents. Bioresour. Technol. 323, 124642. https://doi.org/10.1016/j.biortech.2020.124642 (2021).
CAS Article PubMed Google Scholar
Leng, J. et al. Bi-level artificial intelligence model for risk classification of acute respiratory diseases based on Chinese clinical data. Appl. Intell. https://doi.org/10.1007/s10489-022-03222-y (2022).
Article Google Scholar
Leng, J. & Jiang, P. Granular computing–based development of service process reference models in social manufacturing contexts. Concurr. Eng. 25, 95–107. https://doi.org/10.1177/1063293×16666312 (2017).
Article Google Scholar
Bu, F. & Wang, X. A smart agriculture IoT system based on deep reinforcement learning. Future Gener. Comput. Syst. 99, 500–507. https://doi.org/10.1016/j.future.2019.04.041 (2019).
Article Google Scholar
Hsu, K. L., Gupta, H. V. & Sorooshian, S. Artificial neural network modeling of the rainfall-runoff process. Water Resour. Res. 10, 2517–2530. https://doi.org/10.1029/95wr01955 (1995).
ADS Article Google Scholar
Download references
This work was supported by Chiang Mai University. One of the authors (Y.K.) wishes to acknowledge the CMU Graduate School for Research Assistant Scholarship. We also wish to thank the Department of Industrial Engineering, Excellence Center in Logistics and Supply Chain Management (E-LSCM), Chiang Mai University for providing research facilities. This research is part of the project “Industry 4.0 for SMEs” from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie Grant agreement No 734713.
Chiang Mai University.
Graduate Program in Industrial Engineering, Faculty of Engineering, Chiang Mai University, Chiang Mai, 50200, Thailand
Yotsaphat Kittichotsatsawat
Excellence Centre in Logistics and Supply Chain Management, Chiang Mai University, Chiang Mai, 50200, Thailand
Yotsaphat Kittichotsatsawat & Korrakot Yaibuathet Tippayawong
Department of Mechanical Engineering, Faculty of Engineering, Chiang Mai University, Chiang Mai, 50200, Thailand
Nakorn Tippayawong
Department of Industrial Engineering, Faculty of Engineering, Chiang Mai University, Chiang Mai, 50200, Thailand
Korrakot Yaibuathet Tippayawong
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
Conceptualization, K.Y.T. and N.T.; methodology, Y.K. and K.Y.T.; validation, Y.K. and K.Y.T.; formal analysis, Y.K. and K.Y.T.; investigation, Y.K.; writing—original draft preparation, Y.K.; writing—review and editing, N.T.; supervision, K.Y.T.; funding acquisition, K.Y.T. All authors have read and agreed to the published version of the manuscript.
Correspondence to Yotsaphat Kittichotsatsawat or Korrakot Yaibuathet Tippayawong.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Kittichotsatsawat, Y., Tippayawong, N. & Tippayawong, K.Y. Prediction of arabica coffee production using artificial neural network and multiple linear regression techniques. Sci Rep 12, 14488 (2022). https://doi.org/10.1038/s41598-022-18635-5
Download citation
Received: 30 March 2022
Accepted: 16 August 2022
Published: 25 August 2022
DOI: https://doi.org/10.1038/s41598-022-18635-5
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Advertisement
Advanced search
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.