There is commonly a large volume of data and several types of features in the production process that impact the final product's quality and can be used for process control. Quality control in the production process plays a key role to ensure product quality. So, defect prediction can help quality control and quality management system. Data mining techniques, efficient for deriving complicated relationships in multivariate data-rich environments, can be used to develop automatic defect diagnosis approaches. In this research, using the data of the quality control unit of Bahman Motor Automobile Factory, to predict defects in the production process of single-cab and double-cab chassis products, a hybrid defect prediction model has been developed. Since the raw data available in the industry always contains some misleading, redundant, and contradictory information that must be pre-processed, using techniques such as Random Forest to predict missing data, Mahalanobis to remove outliers, and SMOTE to fix Imbalanced data were used. At first, features were selected using feature selection methods, and Hold-out and K-fold techniques were conducted for cross-validation. K-nearest-neighbor, Bayes-Gaussian, Neural Network, Decision Tree, Random Forest, and Gradient boosting models were evaluated, and finally, a Hybrid model was developed using random forest algorithms and artificial neural network. based on the evaluation metric, this hybrid approach more effectively models the complexity of such an environment compared to the introduced single models. It has a higher capability and shows promise for improving existing quality control approaches, allowing for earlier handling of defects and deviations. As a result, it reduces the amount of waste and costs associated with repairs
