Adapt existing model for Active Learning Fraud Detection -- 2
Budget: €30 – €250 EUR
The project is to adapt this model from GitHub (https://github.com/rmfajri/PS3/blob/master/PS3_final.ipynb), which is an active learning for hate speech recognition on social media, to work with this dataset from Kaggle (https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud).
The big problem is that the code was originally written to work with text, and the script uses text tokenizer. However, I want the code to be adapted to work with the Credit Card dataset, which consists of the features Time, Amount, and V1 - V28 (PCA transformed variables).
When looking at the current code in GitHub, you can see that in [5], there is an iterative for loop. This loop is essential but can be reduced to fewer loops to reduce the processing time.
The big problem is that the code was originally written to work with text, and the script uses text tokenizer. However, I want the code to be adapted to work with the Credit Card dataset, which consists of the features Time, Amount, and V1 - V28 (PCA transformed variables).
When looking at the current code in GitHub, you can see that in [5], there is an iterative for loop. This loop is essential but can be reduced to fewer loops to reduce the processing time.