Data Science Lead Scoring Case Study: Project Overview

• Created a logistic regression model that helps in identifying potential lead conversion rate to organisation
• Optimized model to provide 92% on test accuracy

Code and Resources Used

Python Version: 3.7
Packages: Pandas, numpy, sklearn, matplotlib, seaborn, statsmodel

Data Cleaning:

• Replacing select in data with NaN as these were the category were customer selected nothing which means a NaN object
• Dropping records which has no use for analysis
• Calculated percentage of misiisng values and treated with mean imputation for certain columns
• Inspected numerical columns for outliers and capped to 99 percentile

EDA

I looked at the distribution of data and tried to identify the most important feature in the columns that well explains the leads as well as conversion of customers

• performed Univariate and bivariate Analysis
• found the maximum correlation object with respect to target variable
  

Model Building

First, I transformed the categorical variables into dummy variables. I also split the data into train and test sets with a test size of 30%.

• I started building the model using Stats model and RFE
• PLotted ROC curve and got area of 0.97
• PLotted Accuracy, sensitivity and specificity for various probabilities to find the cut off
  
  
  Here we selected 0.3 as probability cut off

Model Performance

we got final Test accuracies as

• Accuracy: 92.21%
• Sensitivity: 91.78%
• Specificity: 92.48%

MOdel seems to predict the target lead conversion Rate very well