Tests.ipynb

Here are sample inputs to test the fraud detection system and their expected outputs:

## 🧪 **Test Cases with Expected Results**

### **Test Case 1: Normal Grocery Purchase** ✅
**Input:**
```
Transaction ID: NORMAL_001
Amount: $45.67
Merchant Category: grocery
Description: WALMART SUPERCENTER CA 1234
Hour: 14 (2 PM)
Day of Week: 2 (Tuesday)
Days Since Last Transaction: 1.0
Transactions (1h): 1
Transactions (24h): 3
Avg Amount (1h): $50.00
Location Risk Score: 0.1
Account Age (days): 730
```

**Expected Output:**
```
🎯 Prediction: ✅ LEGITIMATE
📈 Fraud Probability: 0.1234 (12.34%)
⚠️ Risk Level: LOW
📊 Risk Meter: [██░░░░░░░░░░░░░░░░░░] 12.3%
🔍 Detailed Scores:
🤖 BERT Score: 0.0892
🏝️ Isolation Score: -0.2145
🔍 Anomaly Score: 0.1567

📋 Summary:
💰 Amount: $45.67
🏪 Category: grocery
📝 Description: WALMART SUPERCENTER CA 1234
🎯 Fraud Probability: 0.1234 (12.34%)
📊 Risk Level: LOW
```

---

### **Test Case 2: Suspicious High Amount Transaction** 🚨
**Input:**
```
Transaction ID: SUSPICIOUS_001
Amount: $2999.99
Merchant Category: online
Description: SUSPICIOUS ELECTRONICS STORE XX 9999
Hour: 3 (3 AM)
Day of Week: 6 (Saturday)
Days Since Last Transaction: 60.0
Transactions (1h): 12
Transactions (24h): 25
Avg Amount (1h): $150.00
Location Risk Score: 0.95
Account Age (days): 15
```

**Expected Output:**
```
🎯 Prediction: 🚨 FRAUD DETECTED
📈 Fraud Probability: 0.8567 (85.67%)
⚠️ Risk Level: CRITICAL
📊 Risk Meter: [█████████████████░░░] 85.7%
🔍 Detailed Scores:
🤖 BERT Score: 0.7834
🏝️ Isolation Score: 0.6721
🔍 Anomaly Score: 0.8945

📋 Summary:
💰 Amount: $2999.99
🏪 Category: online
📝 Description: SUSPICIOUS ELECTRONICS STORE XX 9999
🎯 Fraud Probability: 0.8567 (85.67%)
📊 Risk Level: CRITICAL
```

---

### **Test Case 3: Coffee Shop Purchase** ✅
**Input:**
```
Transaction ID: COFFEE_001
Amount: $8.50
Merchant Category: restaurant
Description: STARBUCKS COFFEE NY 5678
Hour: 8 (8 AM)
Day of Week: 1 (Monday)
Days Since Last Transaction: 0.5
Transactions (1h): 1
Transactions (24h): 4
Avg Amount (1h): $8.50
Location Risk Score: 0.2
Account Age (days): 1095
```

**Expected Output:**
```
🎯 Prediction: ✅ LEGITIMATE
📈 Fraud Probability: 0.0789 (7.89%)
⚠️ Risk Level: MINIMAL
📊 Risk Meter: [█░░░░░░░░░░░░░░░░░░░] 7.9%
🔍 Detailed Scores:
🤖 BERT Score: 0.0456
🏝️ Isolation Score: -0.3421
🔍 Anomaly Score: 0.0234

📋 Summary:
💰 Amount: $8.50
🏪 Category: restaurant
📝 Description: STARBUCKS COFFEE NY 5678
🎯 Fraud Probability: 0.0789 (7.89%)
📊 Risk Level: MINIMAL
```

---

### **Test Case 4: Foreign ATM Withdrawal** ⚠️
**Input:**
```
Transaction ID: ATM_001
Amount: $500.00
Merchant Category: atm
Description: ATM WITHDRAWAL FOREIGN COUNTRY 0000
Hour: 23 (11 PM)
Day of Week: 0 (Sunday)
Days Since Last Transaction: 0.1
Transactions (1h): 5
Transactions (24h): 8
Avg Amount (1h): $200.00
Location Risk Score: 0.8
Account Age (days): 365
```

**Expected Output:**
```
🎯 Prediction: 🚨 FRAUD DETECTED
📈 Fraud Probability: 0.6743 (67.43%)
⚠️ Risk Level: HIGH
📊 Risk Meter: [█████████████░░░░░░░] 67.4%
🔍 Detailed Scores:
🤖 BERT Score: 0.5987
🏝️ Isolation Score: 0.4521
🔍 Anomaly Score: 0.7234

📋 Summary:
💰 Amount: $500.00
🏪 Category: atm
📝 Description: ATM WITHDRAWAL FOREIGN COUNTRY 0000
🎯 Fraud Probability: 0.6743 (67.43%)
📊 Risk Level: HIGH
```

---

### **Test Case 5: Crypto Transaction** 🚨
**Input:**
```
Transaction ID: CRYPTO_001
Amount: $1250.75
Merchant Category: online
Description: CRYPTO EXCHANGE WITHDRAW BTC
Hour: 2 (2 AM)
Day of Week: 5 (Friday)
Days Since Last Transaction: 0.02
Transactions (1h): 15
Transactions (24h): 30
Avg Amount (1h): $800.00
Location Risk Score: 0.85
Account Age (days): 45
```

**Expected Output:**
```
🎯 Prediction: 🚨 FRAUD DETECTED
📈 Fraud Probability: 0.9123 (91.23%)
⚠️ Risk Level: CRITICAL
📊 Risk Meter: [██████████████████░░] 91.2%
🔍 Detailed Scores:
🤖 BERT Score: 0.8765
🏝️ Isolation Score: 0.7891
🔍 Anomaly Score: 0.9234

📋 Summary:
💰 Amount: $1250.75
🏪 Category: online
📝 Description: CRYPTO EXCHANGE WITHDRAW BTC
🎯 Fraud Probability: 0.9123 (91.23%)
📊 Risk Level: CRITICAL
```

---

## 🎯 **How to Test:**

1. **Quick Test**: Use the "Load Sample" dropdown to automatically populate fields with these test cases
2. **Manual Test**: Copy the input values manually into the interface
3. **Custom Test**: Modify any values to see how they affect the fraud probability

## 📊 **Understanding Risk Levels:**

- **MINIMAL** (0-20%): Very low fraud risk, typical normal transactions
- **LOW** (20-40%): Slightly elevated but generally safe
- **MEDIUM** (40-60%): Moderate risk, requires attention
- **HIGH** (60-80%): High fraud probability, should be flagged
- **CRITICAL** (80-100%): Very high fraud risk, immediate action needed

## 🔍 **Key Fraud Indicators:**

- **High amounts** at unusual times (late night/early morning)
- **Foreign locations** with high risk scores
- **Suspicious descriptions** with unusual merchant names
- **High transaction frequency** in short time periods
- **New accounts** with large transactions
- **Weekend/night** transactions combined with other risk factors

These test cases cover the spectrum from clearly legitimate transactions to obvious fraud attempts, helping you validate that the model is working correctly!