JMIR Publications

ABSTRACT

Background:

Diabetes is a public health problem around the world, it is a chronic and incurable disease, although it is a disease that cannot be cured, measures and treatments can be taken to control it and keep the patient stable. Diabetes has been the subject of some research ranging from the prevention of the disease to the use of technologies for its diagnosis and control. Health institutions have information required for the diagnosis of diabetes through various tests and appropriate treatment is provided according to the diagnosis. These institutions have databases with large volumes of information that can be analyzed and used in different applications such as pattern discovery and outcome prediction, to help health personnel in making decisions about treatments or medical prescriptions for diabetes management.

Objective:

To develop a drugs recommendation system for diabetic patients based on collaborative filtering and clustering techniques as a complement to the treatments given by the treating doctor.

Methods:

The dataset to be used contains information from patients with diabetic diseases, available in the UCI Machine Learning Repository. This dataset is analyzed and prepared by applying data mining techniques. Unsupervised learning techniques are used for dimensionality reduction and patient clustering. Drug predictions are obtained using the user-based collaborative filtering approach. The collaborative filtering approach allows creating a patient profile, which is compared with other patients with similar characteristics. Finally, recommendations are made considering the identified patient groups. The performance of the system is evaluated using metrics to assess the quality of the groups and the quality of the predictions and recommendations.

Results:

The principal component analysis technique was used to reduce the dimensionality of the data, as a result eight components best explained the variability of the data. We identified six groups of patients using the clustering algorithm, evenly distributed. These groups were identified based on the available information of diabetic patients, then the variation between these groups was examined to predict a probable medication for the patient. The recommender system achieves good results in the quality of predictions with a value of 0.51 in the Mean Squared Error metric and accuracy in the quality of recommendations of 0.61, which is acceptable.

Conclusions:

This work presents a recommendation system that suggests medications according to drug information and the characteristics of patients with diabetes. Some aspects related to this disease were analyzed based on the dataset used on diabetic patients. The experimental results with clustering and prediction techniques are acceptable for the recommendation process. We believe that our system can provide a novel perspective for healthcare institutions that require technologies that can support healthcare personnel in the management of diabetes treatment and control.