Recognizance is a two Round Machine Learning Competition Organised by the Department of Electrical Engineering IIT BHU(Varanasi) in their Annual Fest Prastuti.
This year the Competition includes
- Round1 - To perdict frauds in electricity bill, More Details - Here
- Round2 - To Detect the images containing the power lines from set of given visible and infrared images. - Here
- Used Feature Aggregation on invoice data
- Used Optuna for Hyperparameter tuning
- Used LGBM Classifier
- Used Preprocessing Using IMAGE DATA GENERATOR
- Trained Deep learning model using stratified KFold
- Finally Used Post processing to increase accuracy from 0.9829 to 0.9895
For more info check codes Round1 Round2
- Best notebook to start
- Optuna Tuning
- LGBM
- Aggreation and Magic Features (Credits - Chris Deotte)
We got 3rd position in Round1, 2nd position in Round2 and overall 2nd position in the event while applying these approaches
Mainak Samanta |
Aman Mishra |