Kids Public

At the 34th Annual Conference of the Japanese Society for Artificial Intelligence, Kids Public made announcements relating to a machine learning technique that can estimate the urgency of a consultation from a preliminary interview in its Pediatrics Online remote healthcare consultation service, which facilitates consultation with a pediatrician from a smartphone.

This initiative aims to ensure that online healthcare consultations can be given in order of urgency. A technique for estimating whether or not the pediatrician will recommend an examination within 24 hours has been developed and tested using preliminary information about the person asking for consultation, based on the natural language of consultation descriptions.

Consultation cases included many instances of data with heavy weighting of information extracted through healthcare consultation exchanges, where it is difficult to estimate the ultimate effects of advice with preliminary information alone. As part of the same research, a technique known as PDLDR has been developed, which enables quite robust model learning even with data sets where there is visible divergence between data used in learning and the assumed targets. In order to determine the precision of the proposed technique, binary classification using multiple techniques was evaluated with the AUC index. While other techniques lacked precision, the proposed technique is reported to have recorded a precision rating of AUC0.74.

Kids Public

The AUC has been successfully improved by means of this technique. On the other hand, precision suffered when dealing with large sizes of data. This is thought to be due mainly to the effects of information other than from preliminary interviews, actual online healthcare consultation records and other factors.

Going forwards, in order to carry out estimations with greater precision, Kids Public believes it will be necessary to find a way of minimizing dependence on conversations from online healthcare consultations through greater systematization to obtain records of preliminary interviews. The company plans to use the experience it has gained as it continues with research to bring about higher-quality online healthcare consultation.