Demand and Revenue Forecasting through Machine Learning

In this Contribution, we investigate and explore on logistics data from sensors and real time devices to derive insights and identify the key features which play deterministic role for predicting the demand and revenue. In this contribution we compare the performance of various time series forecasting models and supervised learning algorithms to predict the demand and revenue to maximize the customer experience and yield greater revenue yield. RMSE has used as Key performance indicator for the comparison, From our analysis results we infer that latitude, longitude, hour of the day and day of week are important in predicting the outcome, further our study indicates Multivariate Time Series forecasting seems to be better performing for revenue and random forest seems to be performing in predicting the demand.