Semenov A., Zelentsov V., Pimanov I. Application Suggesting Attractive Walking Routes for Pedestrians Using an Example of Saint-Petersburg City // Procedia Computer Science, Volume 156, 2019, pp. 319-326. DOI: 10.1016/j.procs.2019.08.208.
Nowadays, when services suggesting the shortest path installed in every smartphone, people use them even if it is not necessary. In most cases when people use vehicles, the shortest path may really be the most relevant, but in case of pedestrians, we are less protected from environment and the factors like noise level, air quality, beauty and safety of route are coming to the fore. Moreover, in most cases we consider several factors deciding which route to choose: we need silent and green at the same time, beautiful as well as illuminated and etc. In most studies this factors are taking into account separately. In our research, using an example of Saint-Petersburg, we developed the prototype application suggesting attractive walking route for pedestrians considering the combination of different factors besides it’s length.
To estimate true score of places from the point of walking attractiveness we conducted a questionnaire, and collected scores of attractiveness in the different places of the city. The acquired samples were used to create model defining attractiveness of place by combination of different factors such as noise, architecture, density of photos and proximity of water bodies as well as green and industrial zones. The model created by Random Forest algorithm classifies places in three groups: attractive, mid-attractive and non-attractive with f1 – scores about 0.6; Using this model, we determined the attractiveness index of all of the footways in the city.
To create navigation algorithm and design mobile application we applied classic Dijkstra’s shortest path finding algorithm with cost weighted by attractiveness index of footways. The preliminary estimation of algorithm shows that provided routes are 14% more attractive and just 6% longer than the shortest ones.