To enhance web application security, this study investigates intrusion detection using a relevant dataset. We introduced a multiclass intrusion detection method, leveraging both machine learning and deep learning, and CIC-IoT-2023 processed the dataset to generate three refined versions. After developing and evaluating several models, including Logistic Regression, XGBoost, Random Forest, SVM, LSTM, and CNN-LSTM, we found that CNN-LSTM achieved superior recall. The application of blending model classification and explanation methodologies leads to improved web application system security and reliability. This study represents a major advancement in web application intrusion detection, delivering a robust and transparent defense against large-scale cyber threats targeting web application systems and their associated data centers.