Chronic Kidney Disease (CKD) is a serious public health concern that requires precise forecasting and early identification to slow its development. A gradual decline in kidney function characterizes CKD. Quality of life can be enhanced, and the course of CKD can be considerably slowed with early identification and care. The traditional methods for the prediction of kidney disease may be prone to error since they involve many clinical parameters to consider. Conventional diagnostic techniques frequently use static data, which can fail to identify temporal patterns that are essential for a precise diagnosis. This research aims to predict the presence of kidney disease by refining the dataset with essential features and stabilizing the features with an equal number of samples. This research proposes Random Oversampled Sequential CNN (ROSC) for predicting the CKD disease with high accuracy by using the CKD dataset from UCI. The novelty of the proposed ROSC model analyzes longitudinal patient data to spot minor changes in features over time that might be signs of kidney malfunction. The first contribution of this research lies in the extraction of the top 10 important features from the dataset after preprocessing. The second contribution is the formation of a balanced dataset by using random oversampling. The important feature extracted from the random oversampled dataset was applied to the existing classifiers and ROSC for assessing the performance. By using the random oversampling, the model training and predicting performance were enhanced by increasing the representation of minority classes. The feature significance analysis used in this research determines the important indicators that have the most impact on the model’s predictions, allowing physicians to concentrate on important risk variables. Implementation results show that the proposed ROSC model predicts the kidney disease with 100% accuracy.