AI Boffins

Timeseries is a sequence taken at successive equally spaced points in time. Rise and fall of stock prices, weather etc. are all different examples of timeseries. Due to its property of persistence, timeseries observations change slowly. Also, past observations provide information about current observations.EON[1] wants to predict a customer’s annual consumption with lesser amount of data as a fairly new technology of smart meters has less historical data. This will help them configure direct debit payments and help in achieving timely payments.The goal is to predict monthly electricity consumption for 3248 households in the coming year (January to December) using historical half hourly energy readings of 3248 smart meters . The half hourly data from EONwas converted to daily data and missing values were imputed. Data was visualized using TSNE and Clustering. Robust STL and Dickey Fuller tests were performed to analyze the seasonality and trend. Due to stationary nature of the data, KNN model was used to obtain the consumption for 3248 meter-ids for the coming year.