In this era of the digital world, we can instantly communicate and share anything on social media with thousands of people in seconds. The data present on social media can be both true and false information. If the information someone shared on social media is false then it will cause social and political harm. So, it is important to spot rumors on the internet to maintain information integrity and to reduce the social impact of misinformation. The motive of this research is to propose a method to detect rumors on social network platforms using the PHEME dataset present on Kaggle. We have compared different machine learning models both with and without hyperparameter optimization and finally proposed a method that archives best accuracy among all. The proposed approach handles the class imbalance using random under sampling and applied pre-processing techniques like tokenization, stemming, and lemmatization. Term Frequency-Inverse Document Frequency (TF-IDF) is used for feature extraction and hyperparameter optimization is performed using the Harris Hawk Optimization (HHO) to retrieve the best hyperparameters. These hyperparameter are then used to train the model using SVM algorithm. Various performance metrics like accuracy, recall, precision, and f1-score are used to evaluate the efficiency of the proposed model. To provide explainability and transparency Explainable AI (XAI) is integrated in the proposed model.