fraud_detector.py

# fraud_detector.py

import torch
import pandas as pd
from transformers import BertTokenizer, BertForSequenceClassification
from torch.utils.data import Dataset
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, precision_recall_fscore_support

class FinancialFraudDataset(Dataset):
    def __init__(self, encodings, labels):
        self.encodings = encodings
        self.labels = labels

    def __len__(self):
        return len(self.labels)

    def __getitem__(self, idx):
        item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
        item["labels"] = torch.tensor(self.labels[idx])
        return item

class FinancialFraudTrainer:
    def __init__(self, data_path=None):
        self.data_path = data_path
        self.tokenizer = None
        self.model = None

    def load_model(self):
        # 從 Hugging Face 模型倉庫載入（或用你訓練好的模型路徑）
        self.model = BertForSequenceClassification.from_pretrained("hfl/chinese-roberta-wwm-ext", num_labels=2)
        self.tokenizer = BertTokenizer.from_pretrained("hfl/chinese-roberta-wwm-ext")
        self.model.eval()

    def predict_transaction(self, text):
        try:
            self.model.eval()
            inputs = self.tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128)
            with torch.no_grad():
                outputs = self.model(**inputs)
                probs = torch.softmax(outputs.logits, dim=1)
                prediction = torch.argmax(probs, dim=1).item()
                confidence = probs[0][prediction].item()
        except Exception as e:
            return f"Error: {str(e)}"