Lead_Scoring

A logistic regression model to assign a lead score between 0 and 100 to each of the leads which can be used by the company to target potential leads. A higher score would mean that the lead is hot, i.e. is most likely to convert whereas a lower score would mean that the lead is cold and will mostly not get converted.

The following are the steps used:

1. Cleaning data: The data was treated for missing values. The values ‘select’ were replaced with null value since it did not give us much information. Columns with more than 40% missing values were dropped, for rest, imputation was done based on their respective count plots. Finally very few rows were left with null values, we dropped those rows.
2. EDA: We performed univariate analysis for categorical and numerical variables. It was found that a lot of categorical variables were not adding any relevant information to the model, therefore, we dropped them. The numeric variables were good and no outliers were found.
3. Dummy Variables: Dummy variables were created for categorical variables excluding binary variables like ‘Do not email’, ‘Do not call’. The original columns for which dummies were created were dropped later.
4. Train-Test split: The split was done at 70% and 30% for train and test data respectively.
5. Scaling: StandardScaler was used to scale numerical variables.
6. Model Building: Firstly, RFE was done to attain the top 15 relevant variables. Few variables were dropped later (manually) depending on the p-value and VIF values.
7. Prediction on train data: We did prediction on the train data using the final model.
8. Model Evaluation(on train data): A confusion matrix was made. We got accuracy of 78.96% , sensitivity of 73.49% and specificity of 82.41%, Precision of 72.49% and Recall of 73.49%.
9. Prediction on test data: Prediction was done on the test data frame.
10. Model Evaluation (on test data): Accuracy of 79.54%, sensitivity of 73.44% and specificity of 82.97%.
11. Lead Scores were calculated, Hot Leads were identified, and Feature importance was identified using res.params.sort_values(ascending=False).
12. Recommendations were provided.