Credit Fraud

By shrijeetverma13 · April 23, 2026

This dataset contains information about credit card transactions, with a focus on identifying fraudulent activity. Each record represents a single…

Fraudulent transactions have a notably higher average amount ($8,404.03) compared to legitimate transactions ($6,604.46). This suggests that fraudsters tend to make larger transactions. A bar chart has been generated to visually compare these two averages side by side.

Customers with a previous fraud flag are significantly more likely to experience fraud again. Those with a prior fraud history have a fraud rate of 91.9%, compared to just 44.3% for customers without a prior fraud flag — making them 2.1x more likely to be defrauded again. A bar chart has been generated to visually compare these fraud rates side by side.

Foreign transactions have a significantly higher fraud rate than domestic ones. The fraud rate for foreign transactions is 77.25%, compared to 43.26% for domestic transactions — a difference of 33.99 percentage points. This means foreign transactions are 1.79x more likely to be fraudulent than domestic ones. A bar chart has been generated to visually compare these two fraud rates side by side.

Yes, accounts with negative balances are disproportionately associated with fraudulent transactions. The bar chart clearly shows that 100% of negative balance accounts are linked to fraud, compared to 48.17% for positive balance accounts — a stark difference that confirms the association.

The bar chart breaks down total fraud losses across different transaction amount ranges, from small transactions under $500 all the way up to $15,000. The visualization clearly shows which dollar ranges are driving the most fraud losses, both in total dollar value and as a percentage of all fraud.

The bar chart reveals clear peak fraud windows throughout the day. Hour 23 (11 PM) has the highest fraud rate at 55.82%, with 710 fraudulent transactions out of 1,272 total. Late-night hours dominate the top fraud windows, suggesting fraudsters are most active when monitoring may be reduced.

The dataset has a notably high overall fraud rate of 48.17%, meaning nearly half of all 32,300 transactions are fraudulent (15,560 fraud cases). Two visualizations were generated: a bar chart showing fraud rate by age group with a dashed reference line at the 48.17% overall average, and a grouped bar chart comparing total transaction counts versus fraudulent counts across age groups.

Yes, there is a slight positive correlation between num transactions today and fraud rates, but it's relatively weak. The correlation coefficient is 0.1227, meaning higher daily transaction counts are mildly associated with increased fraud — but it's not a strong predictor on its own. Two visualizations were generated to illustrate this relationship clearly.

The analysis examined how fraud likelihood varies across different merchant distance ranges (0–5,000 km). Two visualizations were generated: a bar chart showing fraud rate (%) per distance band, and a dual-axis chart overlaying transaction volume with fraud rate. These charts reveal how fraud risk changes as the distance between the customer and merchant increases.

Out of 32,300 total transactions, 6,903 (21.37%) involve merchants with risk scores above 8. This means roughly 1 in 5 transactions are associated with high-risk merchants.

The total monetary exposure from fraudulent transactions is $128,178,209.96, spread across 15,560 fraud cases. On average, each fraudulent transaction results in a loss of $8,404.03. A histogram has been generated showing the distribution of fraudulent transaction amounts, giving you a clear picture of how losses are spread across different transaction sizes.

When foreign transactions combine with a merchant risk score above 10, the fraud rate is 100% — meaning every single transaction in this segment was fraudulent. This is 2.1x higher than the overall dataset fraud rate of 48.17%. A bar chart has been generated comparing fraud rates across four transaction segments for easy comparison.

The analysis examined fraud concentration across different merchant distance ranges for foreign transactions. A bar chart was generated showing both fraud rate (%) and fraud count across 10 distance bands from 0 to 5,000 km, giving a clear visual picture of where fraud is most concentrated.

Among the 15,560 fraudulent transactions, 2,520 (16.2%) had transaction amounts exceeding the account balance, suggesting potential overdraft fraud. In comparison, only 13.0% of non-fraudulent transactions showed the same pattern. This means fraudulent transactions are more likely to involve overdraft situations, though the difference is moderate. A bar chart has been generated showing the breakdown of overdraft vs. normal transactions for both fraud and non-fraud categories.

The analysis examined fraud vulnerability across different age groups, producing detailed data tables that break down fraud rates and average losses by age. Seven data tables were generated showing the distribution of fraud cases, fraud rates (as a percentage of total transactions), and average fraud loss amounts for each age group.

The highest fraud-risk profile combines age 30, transaction hour 13 (1 PM), and a foreign transaction — resulting in a striking 90.9% fraud rate (10 out of 11 transactions). Two visualizations were generated: a bar chart ranking the Top 20 highest fraud-rate profiles by age, hour, and foreign transaction status, and a heatmap showing fraud rates across age groups and transaction hours specifically for foreign transactions. Overall, foreign transactions carry a significantly higher fraud risk than domestic ones, and mid-day hours around 1 PM appear to be a particularly vulnerable window.

Contrary to what you might expect, fraud is NOT higher during nighttime hours. In fact, daytime transactions (6-21) have a slightly higher fraud rate of 49.29% compared to nighttime (22-5) at 45.87%. The difference is statistically significant (p < 0.000001), meaning nighttime fraud is actually lower than daytime fraud.