import pandas as pd import pickle import os import hashlib import json import re import numpy as np import torch from transformers import BertTokenizer, BertForSequenceClassification from flask import Flask, request, jsonify from flask_cors import CORS # from pyngrok import ngrok from sklearn.preprocessing import LabelEncoder from sklearn.feature_extraction.text import TfidfVectorizer from sklearn.metrics.pairwise import linear_kernel from transformers import AutoModel, AutoTokenizer from surprise import Dataset, Reader, SVD from surprise.model_selection import train_test_split from huggingface_hub import login import shutil import os # from huggingface_hub import login os.environ["TRANSFORMERS_CACHE"] = "/tmp/custom_cache" os.makedirs("/tmp/custom_cache", exist_ok=True) # # Get the Hugging Face token from the environment variable hf_token = os.getenv("HF_TOKEN") if hf_token: login(token=hf_token) else: print("Hugging Face token not found. Please set the HF_TOKEN environment variable.") print("Logged in successfully.") # print("Logged in successfully.") print(torch.__version__) app = Flask(__name__) CORS(app) # Determine the device device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print(f"Using device: {device}") # Ensure the file path for the label encoder writable_path = '/tmp/label_encoder.pkl' # Verify /tmp exists and is writable if not os.path.exists('/tmp'): os.makedirs('/tmp') # Define the original labels original_labels = ['usability', 'security', 'both', 'other issue'] label_encoder = LabelEncoder() label_encoder.fit(original_labels) # Save the label encoder with open(writable_path, 'wb') as f: pickle.dump(label_encoder, f) print(f"Label encoder saved to {writable_path}") # Load the label encoder to ensure it's working with open(writable_path, 'rb') as f: label_encoder = pickle.load(f) print("Label encoder loaded successfully") # Load the models and tokenizers sentiment_model = AutoModel.from_pretrained('AliHamza123/sentiment-model').to(device) sentiment_tokenizer = AutoTokenizer.from_pretrained('AliHamza123/sentiment-model') issue_model = AutoModel.from_pretrained('AliHamza123/issue-model').to(device) issue_tokenizer = AutoTokenizer.from_pretrained('AliHamza123/issue-model') subcategory_model = AutoModel.from_pretrained('AliHamza123/subcategory-model').to(device) subcategory_tokenizer = AutoTokenizer.from_pretrained('AliHamza123/subcategory-model') pattern_mapping = { 'Learnability': ['Informative Dialogues', 'User-Friendly Error Messages', 'Conveying Threats & Consequences', 'Human-Computer Interaction for Security (HCI-S)'], 'Memorability': ['Informative Dialogues', 'User-Friendly Error Messages', 'Conveying Threats & Consequences', 'Human-Computer Interaction for Security (HCI-S)'], 'User Error Protection': ['Warn When Unsafe', 'Recoverable Errors', 'Disclose Significant Deviations', 'Security Indicators'], 'Operability': ['Direct Access to UI Components', 'Minimize User Inconvenience', 'Progressive Disclosure', 'Interactive Feedback'], 'Accessibility': ['Direct Access to UI Components', 'User-Friendly Security Policies', 'Contextual Help and Support'], 'Satisfaction': ['Attractive Options', 'Interactive Feedback', 'User Consent for Actions', 'Visual and Auditory Notifications'], 'Efficiency': ['Progressive Disclosure', 'Minimize User Inconvenience', 'System Defaults to Secure Settings'], 'Confidentiality': ['Encryption and Decryption Options', 'Access Control Management', 'Secure Development Practices', 'Security Testing and Evaluation'], 'Integrity': ['Encryption and Decryption Options', 'Authorization', 'Incident Response Plans'], 'Availability': ['Regular Security Updates', 'System Defaults to Secure Settings', 'General Notifications About Security'], 'Authenticity': ['Email-Based Identification and Authentication', 'Multi-Factor Authentication (MFA)', 'Create a Security Lexicon'], 'Accountability': ['Audit and Log User Actions', 'Risk Assessment Tools', 'Migrate and Backup Keys'], 'Non-repudiation': ['Audit and Log User Actions', 'Disclose Significant Deviations', 'Create a Security Lexicon'], 'Traceability': ['Disclose Significant Deviations', 'Audit and Log User Actions', 'Security Features Used by the User'], 'Authorization': ['Access Control Management', 'Authorization', 'Security Features Used by the System'], 'Resilience': ['System Defaults to Secure Settings', 'Regular Security Updates', 'Incident Response Plans'] } # Function to get patterns based on subcategory def get_patterns(subcategory): return pattern_mapping.get(subcategory, []) # Define subcategories usability_columns = ['Learnability', 'Memorability', 'User Error Protection', 'Operability', 'Accessibility', 'Satisfaction', 'Efficiency'] security_columns = ['Confidentiality', 'Integrity', 'Availability', 'Authenticity', 'Accountability', 'Non-repudiation', 'Traceability', 'Authorization', 'Resilience'] def preprocess_text(text): if isinstance(text, str): text = text.lower() text = re.sub(r'[^a-zA-Z0-9\s]', '', text) else: text = '' return text def generate_cache_key(data_dict): data_string = json.dumps(data_dict, sort_keys=True) # Convert data to JSON string, ensuring consistent key order return hashlib.md5(data_string.encode('utf-8')).hexdigest() def get_subcategory(predictions, issue_type): predictions = np.array(predictions) if issue_type == 'usability': max_idx = np.argmax(predictions[:len(usability_columns)]) return usability_columns[max_idx] if predictions[:len(usability_columns)].size > 0 else 'None' elif issue_type == 'security': max_idx = np.argmax(predictions[:len(security_columns):]) return security_columns[max_idx] if predictions[:len(security_columns):].size > 0 else 'None' elif issue_type == 'both': usability_preds = predictions[:len(usability_columns)] security_preds = predictions[:len(security_columns):] usability_subcategory = usability_columns[np.argmax(usability_preds)] if usability_preds.size > 0 else 'None' security_subcategory = security_columns[np.argmax(security_preds)] if security_preds.size > 0 else 'None' return f"{usability_subcategory}, {security_subcategory}" return 'None' cache_dir = '/app/cache' @app.route('/upload', methods=['POST']) def upload_file(): try: # Check if a file is included in the request if 'file' not in request.files: print("No file part in request") return jsonify({"error": "No file part"}), 400 file = request.files['file'] if file.filename == '': print("No selected file") return jsonify({"error": "No selected file"}), 400 # Load the dataset global data if file.filename.endswith('.csv'): data = pd.read_csv(file) elif file.filename.endswith('.xlsx') or file.filename.endswith('.xls'): data = pd.read_excel(file) else: return jsonify({"error": "Unsupported file format"}), 400 # data = data.head(00) data['Apps Reviews'] = data['Apps Reviews'].astype(str).apply(preprocess_text) # Create the cache directory if it doesn't exist if not os.path.exists(cache_dir): os.makedirs(cache_dir) # Generate cache key cache_key = generate_cache_key(data.to_dict()) cache_file = os.path.join(cache_dir, f'{cache_key}.json') # Check if cached file exists if os.path.exists(cache_file): print(f"Cache hit: {cache_file}") with open(cache_file, 'r') as f: result = json.load(f) return jsonify(result) # Define a static user_id for recommendations user_id = 1 # or any appropriate user_id # Processing functions def sentiment_analysis(reviews): inputs = sentiment_tokenizer(reviews, padding=True, truncation=True, max_length=128, return_tensors='pt') input_ids = inputs['input_ids'].to(device) attention_mask = inputs['attention_mask'].to(device) batch_size = 16 all_preds = [] sentiment_model.eval() with torch.no_grad(): for i in range(0, input_ids.size(0), batch_size): end = min(i + batch_size, input_ids.size(0)) batch_input_ids = input_ids[i:end] batch_attention_mask = attention_mask[i:end] outputs = sentiment_model(batch_input_ids, attention_mask=batch_attention_mask) logits = outputs.logits preds = torch.argmax(logits, dim=-1).cpu().numpy() all_preds.extend(preds) return all_preds def issue_identification(reviews): inputs = issue_tokenizer(reviews, padding=True, truncation=True, max_length=128, return_tensors='pt') input_ids = inputs['input_ids'].to(device) attention_mask = inputs['attention_mask'].to(device) batch_size = 16 all_preds = [] issue_model.eval() with torch.no_grad(): for i in range(0, input_ids.size(0), batch_size): end = min(i + batch_size, input_ids.size(0)) batch_input_ids = input_ids[i:end] batch_attention_mask = attention_mask[i:end] outputs = issue_model(batch_input_ids, attention_mask=batch_attention_mask) logits = outputs.logits preds = torch.argmax(logits, dim=-1).cpu().numpy() all_preds.extend(preds) return all_preds def subcategory_classification(reviews): inputs = subcategory_tokenizer(reviews, padding=True, truncation=True, max_length=128, return_tensors='pt') input_ids = inputs['input_ids'].to(device) attention_mask = inputs['attention_mask'].to(device) batch_size = 16 all_preds = [] subcategory_model.eval() with torch.no_grad(): for i in range(0, input_ids.size(0), batch_size): end = min(i + batch_size, input_ids.size(0)) batch_input_ids = input_ids[i:end] batch_attention_mask = attention_mask[i:end] outputs = subcategory_model(batch_input_ids, attention_mask=batch_attention_mask) logits = outputs.logits all_preds.extend(logits.cpu().numpy()) return all_preds # Process the dataset in batches batch_size = 500 # Changed to avoid possible indexing issues with large batches results = [] for start in range(0, len(data), batch_size): end = min(start + batch_size, len(data)) batch = data.iloc[start:end].copy() reviews = batch['Apps Reviews'].tolist() print(f"Processing batch: {start} to {end}") try: sentiment_preds = sentiment_analysis(reviews) issue_preds = issue_identification(reviews) subcategory_preds = subcategory_classification(reviews) sentiment_mapping = {0: 'negative', 1: 'neutral', 2: 'positive'} batch['Sentiment_Result'] = [sentiment_mapping[pred] for pred in sentiment_preds] batch['Identify issue'] = label_encoder.inverse_transform(issue_preds) batch['Subcategory'] = [get_subcategory(pred, issue) for pred, issue in zip(subcategory_preds, batch['Identify issue'])] batch.loc[batch['Identify issue'] == 'other issue', 'Subcategory'] = 'None' # Create the 'Recommendation' column based on 'Subcategory' batch['Recommendation'] = batch['Subcategory'].apply(lambda subcategory: get_patterns(subcategory) if subcategory != 'None' else 'None') # TF-IDF Matrix Calculation tfidf_reviews = TfidfVectorizer(stop_words='english') tfidf_subcategory = TfidfVectorizer(stop_words='english') tfidf_matrix_reviews = tfidf_reviews.fit_transform(batch['Apps Reviews'].fillna('')) tfidf_matrix_subcategory = tfidf_subcategory.fit_transform(batch['Subcategory'].fillna('')) cosine_sim_reviews = linear_kernel(tfidf_matrix_reviews, tfidf_matrix_reviews) cosine_sim_subcategory = linear_kernel(tfidf_matrix_subcategory, tfidf_matrix_subcategory) # Content-Based Recommendation Functions def get_content_based_recommendations_reviews(index, cosine_sim=cosine_sim_reviews): if index >= len(batch): print(f"Index {index} out of bounds for batch size {len(batch)}") return [] sim_scores = list(enumerate(cosine_sim[index])) sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True) sim_scores = sim_scores[1:6] # Get the top 5 most similar items item_indices = [i[0] for i in sim_scores] recommendations = [item for sublist in batch['Recommendation'].iloc[item_indices].tolist() for item in sublist if item != 'No Pattern'] return recommendations def get_content_based_recommendations_subcategory(index): if index >= len(batch): print(f"Index {index} out of bounds for batch size {len(batch)}") return [] try: sim_scores = list(enumerate(cosine_sim_subcategory[index])) sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True) sim_scores = sim_scores[1:6] # Get the top 5 most similar items item_indices = [i[0] for i in sim_scores] recommendations = [item for sublist in batch['Recommendation'].iloc[item_indices].tolist() for item in sublist if item != 'No Pattern'] return recommendations except Exception as e: print(f"Error in get_content_based_recommendations_subcategory: {e}") return [] # Collaborative Filtering interaction_data = { 'user_id': [1, 1, 2, 2, 3, 3, 4, 4, 5, 5], 'item_id': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], 'rating': [5, 3, 4, 2, 5, 1, 3, 4, 2, 1] } interaction_df = pd.DataFrame(interaction_data) reader = Reader(rating_scale=(1, 5)) interaction_dataset = Dataset.load_from_df(interaction_df[['user_id', 'item_id', 'rating']], reader) trainset, testset = train_test_split(interaction_dataset, test_size=0.25) svd = SVD() svd.fit(trainset) def get_collaborative_recommendations(user_id, svd, n=5): try: if user_id not in interaction_df['user_id'].values: print(f"User ID {user_id} not found in the interaction data.") return [] item_ids = interaction_df['item_id'].unique() predictions = [svd.predict(user_id, item_id) for item_id in item_ids] predictions.sort(key=lambda x: x.est, reverse=True) recommended_items = [pred.iid for pred in predictions[:n]] recommendations = [item for sublist in batch['Recommendation'].iloc[recommended_items].tolist() for item in sublist if item != 'No Pattern'] return recommendations except Exception as e: print(f"Error in get_collaborative_recommendations: {e}") return [] def hybrid_recommendations(user_id, index, n=5): content_recs_reviews = get_content_based_recommendations_reviews(index) content_recs_subcategory = get_content_based_recommendations_subcategory(index) collab_recs = get_collaborative_recommendations(user_id, svd, n) hybrid_recs = list(set(content_recs_reviews + content_recs_subcategory + collab_recs)) return [rec for rec in hybrid_recs if rec] # Generate hybrid recommendations for the batch batch['Hybrid_Recommendations'] = [ hybrid_recommendations(user_id, i) if batch['Subcategory'].iloc[i] != 'None' else 'None' for i in range(len(batch)) ] negative_reviews = batch[batch['Sentiment_Result'] == 'negative'] results.extend(negative_reviews[['Apps Reviews', 'Sentiment_Result', 'Identify issue', 'Subcategory', 'Hybrid_Recommendations']].to_dict(orient='records')) except Exception as batch_e: print(f"Error processing batch {start} to {end}: {batch_e}") # Cache the result if not os.path.exists('cache'): os.makedirs('cache') with open(cache_file, 'w') as f: json.dump(results, f) print(f"Total results: {results}") return jsonify(results) except Exception as e: print(f"Error processing request: {e}") return jsonify({'error': 'Internal server error', 'message': str(e)}), 500 if __name__ == '__main__': app.run()