Main Article Content
There are several factors that contribute to distributing infrastructure easily understood losses (NTLs), but since it has a substantial impact on electricity grid performance and value is electricity thefts. Convolutional neural networks (CNNs) and random forests (RFs) are used in this article to support utilities firms address the issues of ineffective power inspections and unpredictable power usage. Convolution and down sampling are used in this approach to teach a convolutional neural network (CNN)  how to distinguish between the various times of day and days of the week in huge and constantly changing data from smart metres. To avoid over fitting, an absent / if these factors is introduced, and the training algorithm technique is used to adjust network parameters during training. Based on these attributes, a random forest (RF) is utilized  to detect if an electrical thief is lurking in the house. RF parameters for the hybrid model are found using the grid search engine. It is concluded that the suggested classification algorithm surpasses existing techniques in terms of effectiveness and precision by carrying out experiments on real datasets.