This paper presents Athena, a predictive approach using traditional Machine Learning (ML) algorithms to mitigate Distributed Denial of Service (DDoS) attacks. The study begins with collecting a dataset containing normal and malicious DDoS packets. After data preprocessing—cleaning, normalization, categorical feature encoding, handling missing values, and data splitting—the project proceeds to feature extraction via ANOVA, identifying the most relevant network traffic features. The model training involves three ML techniques: Random Forest Classifier, K-Nearest Neighbors (KNN), and Long Short-Term Memory (LSTM). The Random Forest model demonstrates the highest accuracy in predicting DDoS attacks. To enhance model interpretability and feature selection, the study employs Local Interpretable Model-Agnostic Explanations (LIME) and ANOVA. The combined use of Random Forest, KNN, and LSTM, supported by LIME and ANOVA, significantly improves network security. Feature importance analysis highlights critical attributes like pktcount and bytecount. The paper concludes by discussing real-time model deployment and the importance of proactive cybersecurity measures.