The purposes of this research is to construct the Cognitive-affective-volitional Integrated Model of Mathematical Learning Process for Junior High School Students（the Cognitive-affective-volitional Model, for short）by using structural equation model in the approach of model generating.  In the first place, the researcher proposed the initial model, which included the latent variables of affection/motivation, volition control, learning strategies, and learning performance.  The research was applied to 447 second-graders in junior high school for the researcher to collect empirical data.  The researcher first tested of the acquired data the hypothesis of multivariate normality distribution with PRELIS 2.50, and then estimated the parameters of the model by means of generally weighted least-squares (WLS) with LISREL 8.50.  The result showed that the initial model could not give a good account of the mathematical learning process for junior high school students, and the statistic results did not seem reasonable, either.  Therefore, the researcher revised the model, and evaluated the parameters of the revised model. After revision, the evaluation results showed that the Cognitive-affective-volitional Model could effectively explain the data.  Though the process of model generating, however, has been revised, the model should be tested with other observation data in future, to see if the model could further be generalized.