An extrapolative model for price prediction of crops using hybrid ensemble learning techniques

Agriculture is the basis for food and the backbone of a country's economy. In India, around 70% of the population is actively involved in growing crops for food or providing direct raw materials to a variety of industries, including textile, food processing, and non-agricultural sectors. The development of technology aids the agricultural sector in forecasting a variety of factors, including crop quality, disease detection, and soil quality, to increase crop yield. However, increased agricultural yield may not always result in a profit due to price reductions. Thus, price forecasting is crucial before choosing the crop to plant since it aids in making informed choices that reduce the risk and loss associated with market price instabilities. This study provides a hybrid model for price prediction that combines an autoregressive integrated moving average (ARIMA) model, a linear statistical analysis for time series data, and an ensemble machine learning approach using support vector regression (SVR). The work has three models: 1) a statistical model is applied over the input features related to crop price, and the residuals are evaluated using SVR, 2) SVR is applied over the predicted price from the statistical model along with the other input features, 3) SVR is applied to the results obtained from the statistical model and its residuals, in addition to the input features. After analysing the results, the model for price forecasting that produced better outcomes was finally chosen. The experimental and result analysis reveals that model 3 has improved results with a 13.37% deviation from actual observation compared to models 1 and 2, which have resulted with deviations of 14.68% and 16.48%, respectively. Additionally, compared to other models, the suggested model has the lowest average prediction errors and average divergence from actual values. Thus, the proposed model is suitable for reliable price forecasting and optimal performance.