Peer Review - pk383

[kothiyalp15]

The proposed project seeks to predict credit score of individuals for the purpose of approving them for a credit card. The rows of the proposed dataset corresponds to each applicant, and has numerous columns regarding their biographical information and financial history. The business use of this project would help lenders and banks minimize credit risk by having a higher accuracy of approving individuals able to maintain their credit obligations, while also benefiting disadvantaged groups who otherwise may not be approved due to internal bias.

Things I liked:

• The dataset has a wide variety of numerical and non-numerical columns; in other words, it is a big messy dataset.
• The scope of this project to New York, JPMorgan Chase data, as well as the fact that this analysis has been a done before, both indicate that this project will definitely be feasible within the timeframe of the semester.
• The project acknowledges evaluating several different models to understand the best in predicting the task.

Areas of Improvement:

• Be cautious of not incorporating too many columns in your prediction to avoid overfitting - initial data analysis may reduce initial feature space
• Consider deploying model across similar data with other regions to ensure applicant's place of origin does not become a factor in your model (for example, California JPMorgan Chase data, etc.)
• Unsure whether approval data is included in the dataset, otherwise maybe specify how the model will be trained?

[kothiyalp15]

Reviewed wrong project, ignore above