Abstract
Drug discovery is a challenging and resource-intensive process characterized by high costs, prolonged development timelines, and regulatory hurdles in the pharmaceutical sector. AI-driven recommendation systems have emerged as an effective approach to enhance candidate selection and optimize drug-target interactions. Typical drug discovery methods are expensive, time-consuming, and frequently have a high failure rate. The inability to quickly identify suitable drug candidates is a significant challenge due to the lack of effective predictive models. To address these issues, the Context-Aware Hybrid Ant Colony Optimized Logistic Forest (CA-HACO-LF) model is proposed. This model combines ant colony optimization for feature selection with logistic forest classification, improving drug-target interaction prediction. By incorporating context-aware learning, the model enhances adaptability and accuracy in drug discovery applications. The research utilized a Kaggle dataset containing over 11,000 drug details. During pre-processing, techniques such as text normalization (lowercasing, punctuation removal, and elimination of numbers and spaces) were applied. Stop word removal and tokenization ensured meaningful feature extraction, while lemmatization refined the word representations to enhance model performance. Feature extraction was further improved using N-grams and Cosine Similarity to assess the semantic proximity of drug descriptions, aiding the model in identifying relevant drug-target interactions and evaluating textual relevance in context. In the classification phase, the CA-HACO-LF model integrates a customized Ant Colony Optimization-based Random Forest (RF) with Logistic Regression (LR) to enhance predictive accuracy in identifying drug-target interactions, leveraging the extracted features and cosine similarity for better performance. The implementation is performed using Python for feature extraction, similarity measurement, and classification. The proposed CA-HACO-LF model outperforms existing methods, demonstrating superior performance across various metrics, including accuracy (0.986%), precision, recall, F1 Score, RMSE, AUC-ROC, MSE, MAE, F2 Score, and Cohen's Kappa.