Scribd Logo
Upload

Welcome to Scribd!

  • 16 views

    Credit Card Fraud Analysis Ashutosh

    Uploaded by

    Hemang Khandelwal
    The document discusses credit card fraud detection using machine learning models. It describes the problem of increasing credit card fraud and the goal of identifying fraudulent transactions. The approach involves data preparation, handling class imbalance, selecting and tuning models like logistic regression, decision trees, XGBoost and evaluating models using ROC AUC score. The best model is selected based on achieving high precision to detect relevant fraudulent transactions or high recall to detect actual high-value frauds depending on the use case.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Credit Card Fraud Analysis Ashutosh

    Uploaded by

    Hemang Khandelwal
    16 views3 pages
    The document discusses credit card fraud detection using machine learning models. It describes the problem of increasing credit card fraud and the goal of identifying fraudulent transactions. The approach involves data preparation, handling class imbalance, selecting and tuning models like logistic regression, decision trees, XGBoost and evaluating models using ROC AUC score. The best model is selected based on achieving high precision to detect relevant fraudulent transactions or high recall to detect actual high-value frauds depending on the use case.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses credit card fraud detection using machine learning models. It describes the problem of increasing credit card fraud and the goal of identifying fraudulent transactions. The approach involves data preparation, handling class imbalance, selecting and tuning models like logistic regression, decision trees, XGBoost and evaluating models using ROC AUC score. The best model is selected based on achieving high precision to detect relevant fraudulent transactions or high recall to detect actual high-value frauds depending on the use case.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    16 views3 pages

    Credit Card Fraud Analysis Ashutosh

    Uploaded by

    Hemang Khandelwal
    The document discusses credit card fraud detection using machine learning models. It describes the problem of increasing credit card fraud and the goal of identifying fraudulent transactions. The approach involves data preparation, handling class imbalance, selecting and tuning models like logistic regression, decision trees, XGBoost and evaluating models using ROC AUC score. The best model is selected based on achieving high precision to detect relevant fraudulent transactions or high recall to detect actual high-value frauds depending on the use case.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    Credit Card Fraud Detection

    Problem Statement:
    Fraudulent activities have increased severalfold, with around 52,304 cases of credit/debit card fraud
    reported in FY'19 alone. Due to this steep increase in banking frauds, it is the need of the hour to
    detect these fraudulent transactions in time in order to help consumers as well as banks, who are
    losing their credit worth each day.

    Every fraudulent credit card transaction that occur is a direct financial loss to the bank as the bank is
    responsible for the fraud transactions as well it also affects the overall customer satisfaction
    adversely.

    The aim of this project is to identify and predict fraudulent credit card transactions using machine
    learning models.

    Approach:
    The approach to the problem can be divided into below parts:

    1. Data Understanding, Data Preparation and EDA

    First look at the data used here from the Kaggle dataset suggests that it is highly imbalanced
    in nature. The positive class (frauds) account for only 0.172% of all transactions:
    Class 0 1
    Count 284315 492

    Class is the target variable which we have to predict where 0 is normal transaction and 1 is
    fraudulent transaction.

    Features V1, V2, … V28 are the principal components obtained with PCA, the only features
    which have not been transformed with PCA are 'Time' and 'Amount'. Feature 'Time' contains
    the seconds elapsed between each transaction and the first transaction in the dataset. The
    feature 'Amount' is the transaction Amount, this feature can be used for example-
    dependant cost-senstive learning.

    Since the PCA transformed variable are already Gausian, there is no need for normalisation.

    We can start with the basic EDA like correlation, boxplots etc for outliers.

    Next, we can use transformation to mitigate and check the skewness in the data. (Boxcox,
    Log transformation, Yeo-Johnson etc)

    Class Imbalances:
    The normal Oversampling method won’t be used here as it does not add any new information to
    the dataset and Undersampling will also not be used as it leads to the loss of information.

    Next, we will try the below two class imbalance handling techniques:

    • SMOTE is a process where you can generate new data points, which lie vectorially between
    two data points that belong to the minority class.

    • ADASYN is similar to SMOTE, with a minor change i.e. the number of synthetic samples that
    it will add will have a density distribution. The aim here is to create synthetic data for
    minority examples that are harder to learn, rather than the easier ones.

    2. Model Selection and Model Building:


    We will start building the model with the train-test split. (At least 100 class 1 rows should be
    there in the test split), use the stratified split here. (80-20 ratio can be used)
    We need to find which ML model works good with the imbalance data and have better
    results on the test data.

    • Logistic regression works best when the data is linearly separable and needs to be
    interpretable.

    • KNN is also highly interpretable, but not preferred when we have a huge amount of data
    as it will consume a lot of computation.

    • The decision tree model is the first choice when we want the output to be intuitive, but
    they tend to overfit if left unchecked.

    • KNN is a simple, supervised machine learning algorithm used for both classification and
    regression tasks. The k value in KNN should be an odd number because you have to take the
    majority vote from the nearest neighbours by breaking the ties.

    • In Gradient Boosted machines/trees. newly added trees are trained to reduce the errors
    (loss function) of earlier models.

    • XGBoost is an extended version of gradient boosting, with additional features like


    regularization and parallel tree learning algorithm for finding the best split.

    We will start with the Logistic regression model with the different value of regularisation and
    hyper param tuning, then go to the decision tree and so on etc. We will also try ensemble
    models and use votingclassifier, Bagging, Boosting etc on the best performing individual
    models to find the best model.

    Hyperparameter Tuning:

    • When the data is imbalanced or less, it is better to use K-Fold Cross Validation for evaluating
    the performance when the data set is randomly split into ‘k’ groups.

    • Stratified K-Fold Cross Validation is an extension of K-Fold cross-validation, in which we


    rearrange the data to ensure that each fold is a good representative of all the strata of the
    data.
    • When you have a small data set, the computation time will be manageable to test out
    different hyperparameter combinations. In this scenario, it is advised to use a grid search.

    • But, with large data sets, it is advised to use a randomized search because the sampling will
    be random and not uniform.

    In our case, we will use the Stratified K-Fold Cross Validation as the dataset is not really
    huge

    3. Model Evaluation:

    We will use sklearn.metrics.roc_auc_score for this as AOC and ROC metric in sklearn is used
    as the metric for highly imbalanced data-set, rest all fails.
    ROC have better false negative than the false positives.
    ROC-Curve = Plot between TPR and FPR
    The threshold with highest value for TPR-FPR on the train set is usually the best cut-off.
    We should not use the confusion matrix as the performance metrics as well as they have
    internally defined hard threshold of 0.5.
    We also can’t completely rely on the precision, recall and F1-score for now as they also
    have their strings attached of some threshold value.
    ROC curve takes into cognizance of all the possible threshold values.
    The ROC curve is used to understand the strength of the model by evaluating the
    performance of the model at all the classification thresholds.
    Because the ROC curve is measured at all thresholds, the best threshold would be one at
    which the TPR is high and FPR is low, i.e., misclassifications are low.

    4. Cost-Benefit Analysis:

    Depending on the use case, we have to account for what we need: high precision or high
    recall.

    For banks with smaller average transaction value, we would want high precision because we
    only want to label relevant transactions as fraudulent. For every transaction that is flagged
    as fraudulent, you can add the human element to verify whether the transaction was done
    by calling the customer. However, when precision is low, such tasks are a burden because
    the human element has to be increased.

    For banks having a larger transaction value, if the recall is low, i.e., it is unable to detect
    transactions that are labelled as non-fraudulent. So consider the losses if the missed
    transaction was a high-value fraudulent one, for e.g., a transaction of $10,000?

    So here, to save banks from high-value fraudulent transactions, we have to focus on a high
    recall in order to detect actual fraudulent transactions.

    We need to determine how much profit or dollar/rupee value we are saving with our best
    selected model.

    You might also like