Real-life data may exhibit skewed data distributions across various classes, which detrimentally influences the learning paradigm. In the two class problems, the number of samples in the majority class substantially exceeds the number of samples from the minority class. It may lead to the marginalization of the minority class by machine learning algorithms, i.e. creating a bias favoring the majority class. To mitigate this issue, the formulation of a strategy to balance them, is vital. Oversampling and undersampling represent two of the principal methodologies for achieving a balanced distribution. This paper proposes a novel approach using Elementary Cellular Automata (ECA) rules to address the data imbalance problem through oversampling of the minority points. Cellular Automata (CA) is a versatile and adaptable method for modeling complex patterns using simple rules and synthetic minority points are generated strategically. CA rule exhibits diverse behaviour, which can be classified into four categories. One representative rule chosen from each category is studied here. This paper shows that it can effectively address imbalances across large and complex datasets. Further, this paper discusses the strengths and limitations of these rules. The proposed technique exhibits accuracy comparable to that of classical techniques, in particular Rule 255 achieves \(>90\%\) accuracy, whereas Rules 110 and 90 achieve \(90\%\) accuracy.