Topic: 'Predicting Red Hat Business Value'
Task: In this task, we identify customers' business potential according to their characteristics and activities to help the company figure out what kinds of people they need to approach and the proper time and ways to approach.
Techniques: Main techniques we are going to use includes PCA for dimension reduction, and for classification we would try several options including Random Forest, Decision Tree, Gradient boosting, Support Vector Machine and Multiple Layer Perceptron. We will evaluate each results to find the optimal algorithm for this senario.
The procedure to conduct our research includes three primary steps:
- Pre-processing and data exploration
- Building models for different algorithms
- Producing ROC curves to evaluate the accuracy of different algorithms