The technique of collecting important knowledge characteristics from a dataset in order to further transform it into usable information is known as data mining. With the combined use of statistics and machine learning, data mining has grown in popularity for a variety of applications, including better decision-making, revenue and operation optimization, cost reduction, anomaly detection, and many more. Despite recognising patterns, One of the most challenging jobs in machine learning is clustering. It is difficult to build an acceptable number of clusters because doing so could decrease the effectiveness of training and assessment. In engineering and scientific applications, the clustering value has steadily increased during the past few years. Many clustering techniques have low classification accuracy, and if we utilise a lot of data with larger dimensions, it will affect the performance and required storage of the algorithms. To reduce this we have to reduce the diamensions of dataset so that clustering algorithm performance will increases and required space will decreases so in this approach we are going to reduce the dimensions of dataset on AFRBFNN [23] algorithm. It is predicated on RL ideas. It categorises every pattern that the two techniques discussed earlier were unable to categorize [23]. Its misclassification rate has decreased. When compared to the other techniques, this model delivers the best classification accuracy. This approach is also quick and has less overlapping. So that we added PCA (Principal Component Analysis) to reduce the dimensions of the dataset and comparing the results with [23] before applying PCA and after applying PCA.