In this study, we try to improve the quality of conventional analysis by applying personal data analysis. To reach the goal, we set two projects. In the first project, we assume latent factors causing disease, and based on the factor, we estimate the health states. For each estimated state, we detect anomaly health data. As a result, the detection accuracy was improved. In the second project, we try to improve the accuracy of the classification of pulse diagnosis. We estimate life patterns from personal health data, and extract health features. By combining the health and pulse features, the classification accuracy was improved. In these projects, we showed that conventional analysis can be improved by considering the personal health data.