Data analytics to increase efficiency of the AI based energy consumption predictor

These days most of the developed countries have been focusing on energy consumption forecasting. The smart management of energy became the key note for many researchers in aspects of better understanding of the complex pattern of energy production, distribution and utility line. The AI based smart predictor mainly depends on the huge historical data analytics along with a set of modelling algorithms to specify each logical part. In this paper we have discussed about the core data analytics and the manual evaluation to capture the exact patterns from a real time versatile building energy consumption dataset which could be utilized in the implementation of an AI based smart predictor. With high dimensional data visualization and data analytics we were able to attain relatively lower mean absolute error i.e. a reduction of error up to 40.22% in the data pre-processing phase. We have mainly focused on the handling of the real time energy consumption data set and to trace out the energy supply or consumption breakdown conditions. From a publicly available data set, various data pre-processing criteria have taken and prepared the data-frame for the multivariate prediction model. After data pre-processing linear regression has been used to predict the energy consumption per carpet area and the prediction efficiency has been compared for each cases. With the manual evaluation and the tracing of energy breakdown or any other unusual condition, the prediction performance increased up to 31.03%.