Since electricity cannot be stored in a large amount, the supply and demand have to be well balanced. As a result, short term load forecast (STLF) would be essential for the power industry to allocate generation resources, spinning reserve, performing system maintenance etc. The objective of the project is to examine the feasibility of machine learning in short term load forecasting. As the electrical loading data in Hong Kong is not opened to the public, the data from Singapore Energy Markey Authority is being used instead. Historical load profile, calendar data and weather data would be used for the forecast. As historical weather forecast is not available, it cannot account for the weather forecast error into the STLF results. Scikit-learn would be used as the tool for machine learning with python as the programming language. Application of such tool would be to provide a more accurate and easy way to perform short term load forecast. If accurate load forecast is available, the power company can better allocate its resources such as lowering the extra generation capacity required to cater forecast error and improve grid security when the forecasted peak load is too high by using the day-ahead demand response scheme.