Since decades prediction of electricity demand has been of utmost interest for better grid management and cost control. An accurate demand prediction can help reduce the emission of greenhouse gases by better planning of extracting energy from cleaner sources. It also allows grids to integrate renewable energy sources like wind and solar energy with non- renewable sources according to the demand. Another advantage of demand prediction is stabilizing the grid by adjusting the generation and distribution according to the forecast as sudden highs and lows in demand can destabilize the grid. This paper explores different machine learning algorithms to predict electricity demand like gradient boosting methods, support vector machine, neural networks, etc. This project focused on comparing different algorithms based on their performance. Traditionally, demand is predicted using statistical methods. An advantage of machine learning models over statistical models is the capability to capture non-linear relationships between data. The data consisted of different hourly weather variables along with the demand. The results showed neural networks and tree-based models performed better than other models of which neural networks were the best. This is because of their capability to handle non-linear data as well as being robust to outliers. The performance of the models was measured using different metrics like R2 score, Mean squared error and root mean squared error.