A wide range of statistical methods are used for the various analyses done in agriculture sector to make today’s agriculture smarter. Technology comprising models has proved to be helpful in getting more production in agriculture and it also can easily handle some of unexpected problems such as floods, draughts and demands during food shortages. Table 2.2 shows the various processes carried out in agriculture sector through analytics and machine learning techniques.
| Authors | Work regarding agriculture | Description of techniques |
|---|---|---|
| Poudel and Shaw | Relationship between climate variability and crop yield | Regression model |
| Rohitashw Kumar and Harendar Raj Gowtham | Impact on climate change and crop productivity | Summary of data |
| Ibn Musah et al. (2018) | Climate change and variability in temperature in Ghana | Continuous distribution |
| Kumar et al. (2018) | Efficient crop yield estimation of sugarcane | SVM, KNN, least square SVM |
| Liakos et al. (2019) | Analysis of soil, water, livestock management | ANN, SVM |
| Priya et al. | Prediction of yield of crop | Random forest |
| Renuka | Analysis of sugarcane crop data of Karnataka state | KNN, SVM, decision trees |
| Zingade et al. (2017) | Android application for suggesting the best profitable crops for the given weather condition | Multiple linear regression |
| Chourasiya et al. (2019) | Seed classification for crop production | ANN, SVM, MLR |
| Veenadhari et al. | Prediction to control climatic parameters for Madhya Pradesh | C4.5 algorithm |
| Abih | Rainfall variability on Ethiopia | Regression, correlation |
| Kumar et al. (2019) | Comparison of weather forecasting using ML algorithms | Decision trees, KNN |
| Kagita et al. | Optimum agriculture land allocation of Krishna delta | Fuzzy membership functions, GA |
| Palanivel and Surianarayanan (2019) | Comparison of crop yield with ML algorithms | Linear regression, ANN |
| Kyei-Mensah et al. (2019) | Analysis of rainfall distribution and crop production for Ghana | Pearson correlation |
| Ndamani and Watanabe (2013) | Inter and intra-annual rainfall variability for crop production in northern Ghana | Coefficient of variation, correlation analysis |
| Kyada et al. | Sensitivity analysis of rainfall forecasting of Gujarat | ANN, adaptive neuro-fuzzy inference system |
| Sumi et al. | Machine learning methods for forecasting daily and monthly rainfall of Fukuoka in Japan | Principal component analysis (PCA), artificial neural network (ANN), multivariate adaptive regression splines (MARS), |
| Peprah (2014) | Correlation among temperature and rainfall of Asunafo forest, Ghana. | Correlation analysis |
| Bewket (2009) | Food grain production and rainfall variability for Amhara region in Ethiopia. | Spearman correlation analysis, coefficient of variation. |
| Shinde and Khadke (2017) | Maharastra’s rainfall on crop production | Standard deviation, Pearson correlation |
| Varsha and Pai | Rainfall prediction of India | Fuzzy C-means clustering, FRBCS |
| Parmar et al. (2017) | Comparison of machine learning algorithms on rainfall prediction | ANN, SOM, CFBP, BPNN, SVM |
| Zaman (2018) | Machine learning model on rainfall for Bangladesh | Regression, decision trees, random forest, KNN |
| Yousif et al. (2018) | Implications of rainfall variability for agricultural production in Eastern Sudan | Coefficient of variation, simple linear regression, correlation |
| Bala Sai Tarun et al. (2019) | Prediction of rainfall with machine learning algorithms | SVM, CART, GA |
| Refonaa et al. | Rainfall forecast of Chennai | Linear regression |
| Arvind et al. (2017) | Analysis of rainfall data for Trichy, Tamilnadu | Standard deviation, CV, probability distribution |
| Lobell et al. | Effects of climate change on crop yield | Regression model |
| Wenjiao et al. | Analysis of crop yields based on climatic contributions | Time series model, cross-section model, panel model |
| Tailor | Crop production of sugarcane in Olpad region of Gujarat | Regression, ANOVA |
| Khatri (2013) | Water irrigation system | Fuzzy logic and artificial intelligence |
| Tzimopoulos et al.(2018) | Relationship between rainfall and altitude of different meteorological stations in Kerala | Fuzzy linear regression |
| Archontoulis et al. | Applications and regression models in agriculture | Nonlinear regression model |
| Menaka and Yuvaraj | Crop yield prediction models | Adaptive Neuro-Fuzzy Inference System, MLR |
Regression is a model that demonstrates the liaison between the dependent and independent variables. Most of the researchers used regression model in their studies in agriculture, while many authors concentrated on data of rainfall and crop production of various regions all around the globe. Regression is one of the supervised machine learning technique which is very flexible and provides high-quality results (Palanivel and Surianarayanan 2019). The rainfall data of different areas of Eastern Sudan have been analysed and it provided reliable R2 values for each station. In Bangladesh, a study of various other algorithms such as naive Bayes, decision trees and regression with 77% of accuracy of results (Zaman 2018) was conducted.
Correlation is another technique of machine learning which is used to evaluate the strapping of two or more variables. In the study of rainfall variability and crop production conducted in Ghana (Ndamani and Watanabe 2013) correlation technique was used to calculate the effect of rainfall (independent variable) on crop yield (dependent variable). In the study of rainfall variability in Ethiopia (Abih 2011), correlation technique was used to state the association between the spring and summer rainfall. Peprah (2014), in the study of Asunafo forest of Ghana, shown the relationship between the climatic conditions of various crops like rice, maize, cassava, cococyam, plantain and yam and rainfall with the help of correlation.
Coefficient of variation is another technique which is used to determine of dispersion of data points in the data sequence around the mean. Ndamani and Watanabe (2013) used coefficient of variation for determining the relationship between the annual and seasonal rainfall of Ghana. Annual rainfall shows coefficient of variation of 0.18, which is taken as moderate rainfall based on correlation and mean. In the study conducted in Ethiopia, Bewket (2009) also used coefficient of variation for comparing the annual, seasonal and daily rainfall levels. Spearman correlation was applied on rainfall data to get its significance. Least square regression was used to best fit the parameter of rainfall.
Continuous distribution is another statistical method which can handle continuous data. When there is a wide range of values, continuous distribution is applied. Some functions such as Weibull, Gumbel and Frechet are the part of extreme value distribution. These functions are used to maximize the maxima value of random variable. GEVD is also a function of continuous distribution which was used in a study done by (Ibn Musah et al. 2018) to get the maximum likelihood values for temperature and rainfall variables.
Probability distribution depicts various possible effects of an incident. It is a mathematical based statistical method. It is divided into continuous probability distribution and discrete probability distribution. This method was used in the study of Tiruchirappalli rainfall data, for evaluating the distribution of rainfall over the region. Mean and standard deviation values are calculated and based on that chi-square values are calculated with various distribution functions. Chi-square values are used to rank the best fit of the distribution (Arvind et al. 2017).
Multiple linear regression (MLR) is modified form of linear regression and it is an effective machine learning technique. Linear regression can handle one independent variable to a dependent variable. But in MLR, more than one independent variables can be assigned to a response variable. Seed classification is done based on soil dataset. Parameters of soil were assigned as independent variables and seed type as a decisive factor in the work of Chourasiya et al. (2019). An android application has been developed by Zingade (2017) with the machine learning technique of MLR. In this application crop prediction as done according to the environmental conditions.
SVM, artificial neural network (ANN), decision trees, random forest and KNN are some supervised machine learning algorithms. Comparative analysis was made with the algorithms such as SVM, ANN, GA and CART on rainfall data. In comparison with other algorithms, ANN provided better performance with 86% accuracy (Bala Sai Tarun et al. 2019). Other algorithms such as SOM, SVM, BPNN, CFBPN and ANN are discussed and reviewed for their benefits in agricultural scenario. The author reviewed various kinds of works regarding agriculture (Parmar et al. 2017).
Leave a Reply