Moe Matsuki, Sozo Inoue, Yoji Kiyota,
In this paper, we analyze the trend and predict the future with the recorded data at a call center which intro- duces caregiving facilities to residents who will be living in. We exploit ensemble machine learning in order to get useful knowledge for three parties, who are “the call center”, “caregiving facilities” and “prospective residents”. Although there are many data analysis research to solve the problems in the caregiving field, they focus on when living in the facilities. In this paper, we analyze the recorded data at a call center to introduce caregiving facilities. The recording data are matrix data of 7,685 rows and 106 columns, with one row for one client, and each column represents the attribute such as “gender” and “degree of care”. In the analysis, first, we changed the explanatory variables to binary. Second, we visualized the importance of attributes and the correlation between positive or negative values for objective variables at the same time by using ensemble learning. As a result, we obtained about 10 findings such that the customers with missing recorded data tends not to contract with a facility, the accesses to facilities are important factor for the decision, the accuracy predicting “whether to observe” achieved 96.8%, and the factor of the decision is not “caregiving service” but “cleaning”, “impression”, “distance”.