In paris-appartemnt-project/apartment_project: Code to Find the Best Apartment to Buy in Paris

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Preprocessing

Before scoring the apartment according to a certain number of caracteristics of the neighbourhood, it is necessary to collect and preprocess the data. The data available in this package is already preprocessed. All the data has been collected from opendataParis except the metro stations information which was gathered from the RATP.

To facilitate further explorations, we divided into two columns (longitude and latitude) the coordinates for each location and convert these coordinates into characters.

This is the data used in the package (in addition to the data scraped from castorus) - cinemas: cinemas in Paris - stations : metro stations in Paris
- commerces: neighbourhood stores in Paris. We only kept Boucherie, Pâtisserie, Tabac,Produits, Alimentaires bio et nature, Alimentation générale, Supermarché, Supérette, Crèmerie - Fromagerie,Poissonnerie, Monoprix, Parapharmacie and Hypermarché from the storesdata. - restaurants: restaurants and bars in Paris. We only keep Restaurant européen, Bar ou Café sans tabac,Restaurant traditionnel français,Restaurant indien, pakistanais et Moyen Orient, Restaurant asiatique, Brasserie - Restauration continue sans tabac, Restauration rapide assise, Restauration rapide debout, Discothèque et club privé,Brasserie - Restauration continue avec tabac,Restaurant central et sud américain, Restaurant maghrébin, Restaurant antillais, Restaurant africain from the stores data. - schools : nursery, elementary and high school in Paris.

The functions used to create the last two databases are commerce(stores), restaurants(stores).

commerce(stores)takes categories of stores that sell food for the everydaylife, such as bakeries, or supermarkets and discards the other types of stores. It then keeps as output their coordinates given by latitude and longitude.

restaurants(stores) filters all stores that have to do with the nightlife. The purpose is to see if the neighborhoods are lively or not given the densities of restaurants, bars...

Ranking the apartments among categories

We then ranked the apartments according to their proximity to a specific criterion. For instance, we compute for each apartment the number of cinemas that are 500m from the apartment and rank them according to the number of cinemas. Then, the apartment with the rank 1 is the apartment that is the best located if we consider that having cinemas nearby is important.

The huge number of restaurants and bars in Paris does not allow us to implement the same process (the time of computations would be to important). Then we decided to divide Paris in 100 clusters using kmeans algorithm. And then compute the number of clusters 500 meters from the apartment and then we associate the number of bars and restaurants in that cluster to the apartment and then rank the apartments according to the number of bars and restaurants. The function to do so is get_rank_restaurants(dataframe,meters)

We proceed in the same way with all the other categories, but we don't need to cluster them as the number of representants for each categories is not too large. Thus, for schools, cinema stations, commerce we apply one universal function : get_rank(dataframe, meters).

Scoring by combination of the rankings.

Then we can join all those rankings with the function join_dataframe(commerces,restaurants,cinemas,schools,stations). To get an index of the livelihood of a neighborhood, we create a composite variable restaurant/cinema. We then classify the apartments in quantiles for each category, thus assigning them numbers from 1 to 10 (1 being the best score).The final step is to combine the scores for each category and give weights to the categories that matter for the user.

paris-appartemnt-project/apartment_project documentation built on May 6, 2019, 4:33 p.m.