As digital data continues to grow exponentially, the reliability of Hard Disk Drives (HDDs) becomes a critical concern for ensuring data integrity and preventing unexpected system downtimes. The research dataset contains historical SMART attribute data collected from HDDs, encompassing parameters such as spin-up time, temperature reallocated sector count and other relevant metrics. The study explored usage of various exploratory data analysis (EDA) & machine learning (ML) techniques for forecasting HDD failures. EDA techniques include Synthetic Minority Over-sampling Technique (SMOTE) algorithm for balancing class representation in training set and Principal Component Analysis (PCA) for dimensionality reduction. Diverse machine learning techniques, such as Random Forest (RF), Decision Tree (DT), K-nearest Neighbors (KNN), Gaussian Naïve Bayes (GNB), Adaptive Boosting (AdaBoost), XGBoost, Gradient Boosting (GB), Linear Discriminant Analysis (LDA), Logistic Regression (LR), are employed to predict the hard drive failure. Local Interpretable Model-agnostic Explanations (LIME) is employed to identify which features have the greatest influence on a specific prediction, providing insights into the factors driving the model’s decisions. Random Forest exhibited superior performance of accuracy 99.9 and F1 score 0.98 compared to other models.