Objective: This paper addresses the issue of model inputs selection before the forecasting exercise. Appropriate data analysis is one of the basic steps in obtaining accurate load forecast. It shapes the forecasting data in to working data by reducing the variation between the individual forecasting variables, or reduces the number of the model inputs. Also, the information received from data analysis determines the method to be used, or how to use it. Methods/Statistical Analysis: It employs the use of correlation analysis to select the forecasting variables, and wavelet transforms to decompose the selected data in to a number of approximations. The purpose is to select the actual forecasting variables, and to limit the variation between them (model inputs). ANFIS was used to forecast the load using the processed data. Findings: From the result obtained, it was observed that selecting the data based on correlation analysis, and wavelet transform improve the accuracy of the forecast, and enhanced the forecasting speed. Applications/Improvements: Improving the forecasting accuracy will save the utility economically, and improving the speed will enhance the time taken to make crucial decisions in power system operation