Azure Revenue Forecasting using Ensemble learning
The goal of this project was to build a model to predict Azure revenue for every customer for next 6 months. _x000D_ A hypothesis driven approach was taken to identify the factors which could be useful predictors. The factors were further grouped into Customer attributes, Usage patterns, spending patterns. Over 60 features are fed into the model. _x000D_ The model architecture is divided into two layers. At Level 0, there are four families of base learners GBM, Random Forest, Neural Net / Deep Learning and Generalized Linear Model. At level 1, a Generalized Linear model takes the outputs of base learners as inputs and gives the final output. The accuracy metric used as performance measure is Mean absolute error and is tracked month over month.