Stacked Ensembles: Boosting Model Performance to New Heights Based on Regression for Forecasting Future Wheat Commodities Prices in Gujarat

Abstract

Objectives:The goal of this study is to develop an estimation method that will improve absolute and proportionate price predictions, helping farmers in their long-term efforts to boost output and profit. Methods: For the experimentation, Dataset is made up of wheat commodities price of different mandi's of Gujarat region which is collected from agmarknet website, run by Indian government and weather variables which are obtained from weather API (world weather online website) based on commodities region such as Max-Min Temp, Pressure, Wind, WindSpeed, Humidity respectively which has a direct impact on crop production. Various machine learning approaches such as SVR (Support vector regressor), DTR (Decision tree regression), PR(Polynomial regression), Multiple regression, and Boosting algorithms such as XgBoost, Light GBM experimented for Daily data, 10 days and 15 days average price estimation, from the experimentation, observed that DTR Can't be performed well with continuous numerical data. The presence of one or two outliers in the data can seriously affect the results of PR (sensitive to the outliers). The meta-learner attempts to minimize the weakness and maximize the strengths of every individual model such as Multiple regression and SVR (Support vector regressor). Findings: The development of a stacking model, which aids a farmer in more accurately predicting crop price, is motivated by the study of several agriculture-specific requirements like weather parameters, and fuel cost. Every individual regression model like the Decision tree regressor, Support vector regressor, and polynomial regressor is evaluated by the meta-learner to identify its strengths and weaknesses. Novelty: This research addresses the complexity of crop price forecasting along with unique feature set which includes (i) historical commodities prices (ii) weather data (iii) transportation factors iiii) Holiday Impacts.