Electrocardiogram (ECG) arrhythmia classification seeks more attention in the research areas for preventing and diagnosing cardiovascular diseases. The conventional approaches related to arrhythmia classification have achieved reasonable performance on the detection of diverse heart scenarios particularly when processing with the imbalanced datasets. Recently, deep learning shows an enhanced performance in the area of healthcare industry for extracting the high level of abstract features in an automatic way to reduce time consumption and avoid manual effort. Many researchers have developed deep learning-aided methods for detecting the arrhythmia with the help of ECG signals; however, it fails to ensure the accurate classification. Hence, this paper plans to develop the arrhythmia classification using ECG signals by the intelligent technology. Initially, the data is gathered from online sources. The gathered data undergo preprocessing using noise removal, artifacts removal, and peak detection techniques. From the pre processed signals, the features are extracted using the “Mel Frequency Cepstral Coefficients (MFCC), spectral features, and Short Time Fourier Transform (STFT)”. As the extracted feature length is lengthy, significant features are extracted in the optimal feature extraction step using the Squirrel Search Algorithm (SSA). These optimally extracted features are given to the final classification step, where “Bi-directional Long Short Term Memory (Bi LSTM)” classify the heartbeats in a more efficient and accurate manner into 5 prominent classes such as “Normal Sinus Rhythm (N), Left Bundle Branch Block (LBBB or L), Right Bundle Branch Block (RBBB or R), Premature Ventricular Contraction (V), and Atrial Premature Beat (A)” that determines the types of arrhythmia present. The comparison illustrates the success of the proposed ECG classification model.