A New Modified Approach of Linear Regression and Decision Tree Modeling for Enhancement of the Accuracy

Forecasting of electricity demand poses various challenges, from ensuring data quality and dealing with complexity and variability. Electrical load forecasting was performed earlier using the conventional method, ANNs which were used for electrical load prediction, have been found to possess certain limitations. As a result, researchers have turned towards alternative approaches, such as Machine Learning (ML) to enhance the forecasting accuracy, and precision and tackle specific challenges. In current days, the selection of emerging and best-fitted technologies, and modeling methods involves a setback in electrical load forecasting due to its growing complexity. After extensive study of various tools and techniques, it has been observed that the Linear Regression Model (LRM) and Decision Tree Model (DTM) are considered the best models for electrical load forecasting. However, it has been observed that inefficient data-keeping systems unknowingly introduce some malicious data into the dataset. Contributing to this, authors have proposed a modified training and testing algorithm for LRM and DTM by incorporating two additional functions. This approach will enhance the data quality and will help in efficient training and testing of the tools. Analysis have shown significantly enhanced accuracy of the prediction. To validate the LRM and DTM modified algorithm’s accuracy authors have tested both the models and compared them. After repeated training and testing, it has been observed that the proposed modified algorithms have shown approximately 10% more accuracy as compared to the results derived from the existing algorithm.