Repository
Presentation

Scoring Banking Leads

Miguel Chacón

Content

• Project Description
• Hypotheses / Questions
• Dataset
• Analysis
• Cleaning
• Model Training and Evaluation
• Deliverables
• Links

Project Description

In this project, I use Machine Learning with Python to create a lead scoring model, leading a 50% increase in conversion rate. It is meant to show my understanding of:

1. Machine Learning concepts, pipelines and model building & tuning.
2. Digital Marketing and conversion rate optimization.
3. Python, Pandas, and Scikit-learn.
4. Modeling and forecasting of business processes.

Business Question

The client, a commercial bank, wants to improve its sales without hiring more sales representatives. I use Business Analytics frameworks to approach the question and deliver a solution.

The lead scoring model I delivered reduces the proportion of leads that are contacted while simultaneously increasing the conversion rate. All else equal, this will allow further lead-generation campaigns to increase their lead volume, leading to an increase in revenue and, thus, profitability.

Dataset

The dataset consists of 45,000 instances of phone calls made by sales representatives during a sales campaign.

Analysis

• Overview the general steps you went through to analyze your data in order to test your hypothesis.
• Document each step of your data exploration and analysis.
• Include charts to demonstrate the effect of your work.
• If you used Machine Learning in your final project, describe your feature selection process.

Cleaning

1. Dimensionality reduction achieved by merging categories.
2. Columns with multicollinearity were dropped.
3. Categorical columns were encoded.
4. The target variable was changed to a numerical type.

Model Training and Evaluation

1. Tested 46 combinations of 8 algorithms and 6 scalers with default settings to find the combination with the highest recall score.
2. Selected a Random Forest Classifier and tested 100 parameter combinations with a 3-fold cross-validation for each, optimizing for Recall. The resulting model had 65% higher score than the model with default settings.
3. To evaluate the model, I took a random sample of 15 leads and made educated guesses on whether they would convert, comparing my results against the model. The model's predictions led to a 50% higher conversion rate.

Deliverables

1. A Machine Learning model that allows new leads to be assigned a score between 1 and 10 based on their likelihood to convert to sales, allowing the sales team to prioritize these for calls.
2. A PPT presentation detailing the business question, the model as a solution, and the results.

Links

Repository
Presentation