Fraudulent_transactions_Detection

This case requires to develop a model for predicting fraudulent transactions for a financial company and use insights from the model to develop an actionable plan

1 . Data cleaning including missing values, outliers and multi-collinearity.

• There is no null values in any columns
• Most of the datapoints are not correlated
• Oldbalance is correlated with newbalance of person who intitiated transaction
• newbalanceDest and oldbalanceDest is also correlated.

2. Describe your fraud detection model in elaboration.

Fraud Detection Model contains following steps:

• Loading of dataset and Explore it
• Checking the Distribution of transaction datatype
• Checking for the need of data cleaning
• statistical Analysis of Data
• Deal with the skewed data with boxcox tranformation and log transformation
• Taking features to train the data
• Train test split with 80-20 rule
• use Sampling to balance the dataset.
• Fit the model to RandomForestClassifier and Logistic Regression
• Analyse the results

3. What are the key factors that predict fraudulent customer?

'isFlaggedFraud', 'oldbalanceOrg', 'type_CASH_OUT', 'type_TRANSFER' are fraud causing/predicting features.

4. Do these factors make sense? If yes, How? If not, How not?

Yes, these factors absolutely makes sense.

• 'isFlaggedFraud' - column which marks transaction as a fraud or not

• 'oldbalanceOrg' - indicates that customer/individual with more balance in his account is prone to fraud transaction

• 'Cash-Out' - it refers to convert non-cash asset into Cash. Thus online selling of large goods etc is more prone to frauds.

• 'Transfer' - Uninformed transfers mode have found to be modes through which fraud takes place.

  5. What kind of prevention should be adopted while company update its infrastructure?

  • Can implement multiple account to distribute the huge balance.
  • Sperate account for online transactions.
  • Special mode for cash out transactions.
  • Special regestering of account users for transfer modes.

  6. Demonstrate the performance of the model by using best set of tools

• Accuracy is used for measuring accuracy.

• Only accuracy cannot be relied as in Fraud detection problem, recall and precision is more important than other metrics.

• We Get highest Precision in Logistic Regression

• highest Accuracy in Random Forest Classifier

• highest Recall in Random Forest Classifier

• highest F1 Score in Random Forest Classifier

By this I concludes Random Forest Classifier is best model for this fraud transaction detection