Physical conditioning has evolved into a cornerstone of athletic preparation. An increasingly hot subject in the field of intelligent sports training technology is the application of AI analysis methods to the problem of determining the efficacy of physical training for athletes. One potential solution is to build an analysis model that integrates optimization of training parameters with data analysis tools. Researching evaluation models for the effects of physical training is crucial for maximizing the benefits of such training for athletes. It is possible to create relevant individualized training programs based on the assessment outcomes. There has been a lot of recent success using neural networks in sports training as a large data analysis method. An improved ant colony optimization—back propagation (IACO-BP) approach for evaluating the effectiveness of physical exercise and creating individualized training programs is proposed in this study, which integrates artificial neural networks with assessment techniques. The primary goal of this study is to develop an improved ant colony optimization algorithm (IACO) that incorporates local route optimization and differential initialization pheromone. It improves the likelihood of achieving the ideal parameters of the prediction model and replaces the back propagation (BP) network's own gradient descent approach for optimizing parameters. The second contribution of this study is the development of an IACO-BP physical training impact evaluation model that uses the enhanced ACO optimized BP network to enhance the reliability of the evaluation. Third, this study validated and confirmed the dependability of the approach by conducting a number of systematic tests on IACO-BP.