Hybrid MLR-ANN model for crop yield prediction

Overview

This solution offers a hybrid predictive model that combines Multiple Linear Regression (MLR) with an Artificial Neural Network (ANN) to improve crop yield forecasting accuracy. Traditional ANN models rely on random initialization of input layer weights and bias, which can lead to slower convergence and less accurate predictions. By using MLR-derived coefficients and intercept to initialize the ANN's input layer weights and bias, the hybrid MLR-ANN model leverages the statistical strengths of linear regression to guide neural network training, resulting in more accurate and efficient yield predictions for crops such as paddy.

Technical specifications

• Hybrid architecture: Combines MLR and Feed Forward ANN with Back Propagation training algorithm
• Smart initialization: MLR intercept and coefficients are used to initialize the ANN's input layer weights and bias, replacing random initialization
• Comparative validation: Prediction accuracy benchmarked against standalone ANN, MLR, Support Vector Regression (SVR), k-Nearest Neighbour (KNN), and Random Forest (RF) models
• Performance evaluation: Uses standard performance metrics and computational time analysis to validate efficiency gains
• Target application: Designed for crop yield prediction, with demonstrated focus on paddy crop yield forecasting

Technology readiness level

The hybrid MLR-ANN model has been developed and validated through comparative analysis against multiple conventional machine learning and statistical models. Results demonstrate that the proposed hybrid model delivers better prediction accuracy than standalone ANN, MLR, SVR, KNN, and RF approaches. The model is ready for further field validation and pilot deployment in agricultural yield forecasting applications.