Solution to Traveling Freelance Teacher Problem using the Simple K-Means Clustering

Abstract

This article proposes viable solutions to the traveling freelance teacher problem, which is similar to the traveling salesman with the time window problem. Freelance teachers are usually paid higher than fully employed teachers, but they are often offered fewer hours and a less consistent schedule. However, freelancers can work at more than one private school, private students, and charity-based work at a local hall or religious places such as a mosque. Mosques in Indonesia are located both in urban areas, where many mosques are nearby and in remote villages, sparsely populated rural areas, where the travel time between mosques is significant. The traveling teacher problem is solved by clustering teaching locations using the K-means clustering approach, and clusters are joined to end the route with the shortest distance. This study used the K-means to automatically search for unknown clusters of learning centers similar to a teacher's geolocation. Results show that using K-Means, the approach can recommend a similar learning center to the teacher's requirement at a cluster number index of about 83%. This study also revealed that the best number of clusters was determined by cluster number index, cluster sum of squared errors, silhouette, and time taken.