Dataset Overview

Basic Information

• Dataset Name: Fund Switches Model ML Web Application
• Source: Created by data experts and subject‑matter experts
• Record Count: 460,000
• Feature Count: 120
• Characteristics: Highly imbalanced; few instances of fund‑manager switching

Problem Description

• Prediction Goal: Forecast fund managers who are likely to switch to another fund
• Main Challenges:
  • Severe class imbalance
  • Complexity of financial data
  • Very few observed switching events in the industry

Solution

• Model Type: Ensemble model (VotingClassifier)
• Base Classifiers:
  • RandomForestClassifier
  • XGBClassifier
  • LightGBMClassifier
• Primary Evaluation Metric: Recall

Model Performance Metrics

• Accuracy: 97.62 %
• Precision: 62.66 %
• Recall: 65.88 %
• F1‑Score: 64.23 %
• ROC‑AUC: 94.9 %

Related Technologies

• Programming Language: Python
• Key Libraries:
  • scikit‑learn
  • pandas
  • numpy
  • openpyxl
  • scipy
  • xgboost
  • lightgbm
  • Flask

Application Demonstration

• Web‑App Framework: Flask
• Main Features:
  • Raw dataset upload
  • Internal training, preprocessing, and validation
  • Display of model metric results
  • Download of prediction results as an Excel file

Future Improvement Directions

• Integrate additional models into the UI
• Enhance user experience
• Deploy at scale using Django

Contact Information

• Name: Mohammed Aftab
• Email: maftab@convergenceinc.com
• Organization: Invergence Analytics