One of the key reasons for cancer-related deaths is lung cancer, which stresses the importance of early detection and accurate predictive models worldwide. This work uses deep learning methods on clinical data to present a predictive healthcare system for lung cancer risk assessment. We expanded the data to 449 cases using the UCI lung cancer dataset, which consists of 16 symptoms and risk variables, therefore guaranteeing a strong analysis. Among several machine learning models used were XGBoost, random forest, decision trees, and a suggested Long Short-Term Memory (LSTM) network. Our aim was to solve overfitting by collecting complex patterns in patient data, hence improving the accuracy and dependability of lung cancer forecasts. Accuracy, precision, recall, and F1-score defined the evaluations of the models. With a training accuracy of 99%, testing accuracy of 98%, and validation accuracy of 98%, the LSTM network out of all the models tested attained the best performance. With better generalizing power and the capacity to control complicated temporal relationships in clinical data, the proposed LSTM model routinely exceeded conventional models. BiLSTM’s incorporation enhanced the classification accuracy even further, and hence, it is a perfect tool for lung cancer prediction applications.