The vast amount of information in a medical database makes data classification a difficult challenge in data mining. When used to medical data, associative classification improves classification accuracy and disease prediction. With this goal in mind, the proposed study presents three methods for accurately categorizing patient medical data via the creation of optimal association rules. The three methods are Logistic Fully Recurrent Deep Neural Learning Classification (LFRDNLC), Adaptive Bio-Inspired Gene Optimization Based Deep Neural Associative Classification (ABGO-DNAC), and Gene Optimized Association Rule Generation based Integral Derivative Gradient Boost Classification (GOARG-IDGBC). The improved results for diabetic illness diagnosis with higher classification accuracy and less time consumption are produced by using the aforementioned three recommended methodologies. The suggested GOARG-IDGBC method's primary objective is to boost classification precision while diagnosing diabetes. The fitness function of each characteristic is evaluated as part of an optimized evolutionary algorithm to generate the best possible set of association rules. The suggested GOARG-IDGBC method employs an integral derivative gradient boost classifier to perform classification based on previously specified association rules (IDGBC). Attributes are categorized using a decision tree in IDGBC.