Data Science project: Analyzing Automobile dealership data.

This project aims in predicting the probability of capturing certain deals based on customer information and vehicle information.

---

Description:

Classes marked as 1 depict the deals that are captured and observations with class marked as 0 depict that the deals are not captured. The dataset is an imbalanced with majority of observations belonging to uncaptured deals (Class = 0). Also it contained few missing values and outliers which were dealth with during pre-processing stage.

Explanations:

The result displays the stages of data preprocessing as well as implementation of Machine learning algorithm (Ensemble methods).
  - Performed Feature Selection to remove all irrelevant features from the dataset.
  - After the dimension is reduced, used IQR (Inter Quartile Range to identify and remove the outliers.
  - Used Bagging classifier technique (Ensemble method) to classify the data and further predict the probability of occurence of Class = 1
  - Bagging classifier provides good results with least misclassified observations for my imbalanced dataset.
  - Calculated f1_score, area under curve and obtained confusion matrix to evaluate the working of my model.
Original percentage of deal being captured is around 5.8%. My results come to around 6.3%.

Programming tools and Packages:

I used PyCharm IDE with Anaconda 3 as project interpreter.
  -> Anaconda 3
  -> scikit-learn
  -> pandas
  -> numpy

---

Results:

  - f1_score: 0.905041605482
  - Confusion Matrix:
    0   1
0   71809   6
1   770   3698
  - Area under curve: 0.91378991803
  - Average probability of deals being Captured is: 0.0631150452918710