The Electrocardiogram (ECG) non-invasively records heart electrical impulses revealing Cardiovascular diseases (CVDs) and intracardiac conducting tissue performance. Globally 17.9 million people died from CVDs in 2019, accounting for 32% of all deaths and further the prevalence will double by 2030. Mostly CVDs are reported in the last stage, thus limiting the choices of methodological diagnosis to the healthcare fraternity. Furthermore, there is a dearth of morphologically based classifications for atrial arrhythmias. This study analyses, classifies, and predicts heart rhythms, atrial disorders, and atrial arrhythmias using Artificial Intelligence (AI) techniques like Machine Learning (ML) and Deep Learning (DL) based on ECG signal morphological features like peaks, waves, and intervals. Initially, this study examined Heart Rate Variability (HRV) in Sinus Rhythm (SR) and Exercise-induced Sinus Tachycardia (ST) in two phases. First, time and frequency domain characteristics of HRV and morphological traits like P-wave and PP Interval were statistically analyzed. Compared to SR and ST, HRV durational properties changed significantly. The subsequent phase of the study entails creating a Long Short Term Memory (LSTM)-based Recurrent Neural Network (RNN) to forecast ST condition heart rate. Further, the heart rate is identical in ST and Atrial Tachycardia (AT), making it challenging to distinguish manually based on the ECG signal. The three ML methods utilized to classify SR, ST, and AT were Extra Trees (ET), Ridge Classifier (RC), and CatBoost (CB).The classification approach uses ECG signal morphological properties. Subsequently, these features were prioritized based on their significance in distinguishing SR, ST, and AT, and the atrial features obtained priority. Later, the work analyses and classifies AT and Left Atrial Enlargement (LAE), AF antecedents, where the morphological features were P Wave Indices (PWI), temporal, and amplitude aspects of the ECG signal. The first half classified SR, ST, AT, and LAE using stacked ML models, and the second half ranked morphological features using a pie method where PWI-based features obtained the highest importance. Finally, this study distinguished Paroxysmal AF (PAF) and Persistent AF (PsAF) from Non-AF cardiac rhythm. Classifying AF subtypes and Non-AF is crucial to improve clinical decision management and defining AF&rsquos clinical status. A 2D custom Convolutional Neural Network (CNN) model automatically classifies AF subtypes based on a time-frequency spectrum where Constant Q Transform (CQT) converted ECG signals to a time-frequency spectrum. Medical professionals may find it helpful to use morphological feature-based ML models to get insight into crucial clinical ECG features for early atrial arrhythmia prediction. Additionally, the generated DL model may help medical professionals choose individualized AF treatment and decrease misdiagnosis.