Analyzing obesity-related factors and classifying individuals based on their risk status are critical for improving the effectiveness of health policies. This study aimed to predict individuals’ predisposition to obesity using classification algorithms. The dataset, obtained from an open-access data source, contains 22,869 records and 16 attributes, including demographic information, physical activities, and dietary habits. Various preprocessing steps were performed on the data to improve model accuracy and reliability by improving data quality. This preprocessing phase included removing missing data, scaling data, and processing outliers. Principal component analysis and linear discriminant analysis were used to identify the most useful attributes by removing irrelevant attributes from the data. After identifying useful features in the dataset, a hybrid model was developed to classify individuals’ obesity levels by leveraging the strengths of various machine learning (support vector machines, decision trees, gradient boosting, random forest, and logistic regression) and deep learning algorithms (autoencoders, convolutional neural networks, and recurrent neural networks), as well as individual classification algorithms. In order to achieve higher classification performance, a hybrid model was developed. K-Fold cross-validation was performed to evaluate the performance of this model alone. The hybrid model demonstrated the highest classification performance among all individual models, with an accuracy rate of 90%. Considering the global impact of obesity, the findings of this study will make significant contributions to many areas such as early diagnosis and risk assessment, development of health policies, reduction of health expenditures, efficient use of resources, and development of personalized intervention strategies.