India is seeing an explosion of new competitors in the Broadband space. 'India Broadband' is a company that is now seeing a lot of customer churn due to customer dissatisfaction because of broadband outages.
The company has now curated a dataset, where it tracks several variables that it believes impact the outage_duration. They have tracked three different outage durations, 0 for no outage, 1 for short outages that last anywhere between a few minutes and a maximum of 2 hours, and 2 for long outages that can last from 2 hours to sometimes even a couple of days.
We will now have to use the metrics that the company has tracked to create a machine learning model that will be able to predict the outage_duration so that the company can better handle outages and improve customer satisfaction and therefore reduce customer churn.
- numpy
- pandas
- matplotlib
- seaborn
- functools
- sklearn
- xgboost
- imblearn
- torch
- pytorchtools
- Basic EDA was performed on the dataset
- Data Wrangling performed to aid model building
- Answered the questions using visualizations
- Found class imbalance in the train dataset
- Merging of the datasets was performed
- Train and Validation split
- Pipeline for preprocessing categorical and numerical features
| Model | Train Score | Validation Score | Test Score |
|---|---|---|---|
| Logistic Regression | 75% | 67% | 57% |
| KNN Classifier | 79% | 69% | 58% |
| SVM Classifier | 88% | 70% | 59% |
| Random Forest | 97% | 76% | 65% |
| XGBoost | 90% | 77% | 65% |
- Merging of the datasets was performed
- Dropped the area_code as a feature
- Upsampling the non-majority classes
- Train and Validation split
- Pipeline for preprocessing categorical and numerical features
| Model | Train Score | Validation Score | Test Score |
|---|---|---|---|
| Logistic Regression | 69% | 67% | 57% |
| KNN Classifier | 83% | 78% | 59% |
| SVM Classifier | 66% | 67% | 60% |
| Bagged SVM Classifier | 66% | 67% | 62% |
| Random Forest | 86% | 82% | 65% |
| XGBoost | 91% | 87% | 66% |
- Merging of the datasets was performed
- Dropped the area_code as a feature
- Upsampling the non-majority classes
- Train and Validation split
- Pipeline for preprocessing categorical and numerical features
- Attempt 1 -> 63% on test
- Attempt 2 -> 66% on test
- Attempt 3 -> 65% on test