Refactored geocoding.

Geocoding was failing due to input names were written in Spanish but official addresses are in Catalan. This commit introduces a previous step that "translates" addresses from Spanish to Catalan.

.gitignore

+.Rproj.user
+.Rhistory
+.RData
+.Ruserdata

scraping/geocoding-viviendas-turisticas-comunitat-valenciana.R

+# require(devtools)
+# devtools::install_github(repo = 'rCarto/photon')
+library(photon)
+library(stringdist)
+library(tidyverse)
+
+
+# Addresses' cleanup ------------------------------------------------------
+
+# Load Valencia's official street names from:
+# http://gobiernoabierto.valencia.es/va/dataset/?id=listado-de-calles
+calles_valencia = read.csv("data/original/vias-valencia.csv") %>%
+  mutate(codtipovia = sub("C/", "C", codtipovia)) %>%
+  mutate(tipovia_es = codtipovia) %>%
+  mutate(tipovia_es = sub("PG", "PA", tipovia_es)) %>%
+  mutate(tipovia_ca = codtipovia) %>%
+  mutate(tipovia_ca = sub("C/", "Carrer de", tipovia_ca)) %>%
+  mutate(nombre_es = paste(tipovia_es, traducnooficial, sep = " ")) %>%
+  mutate(nombre_es = as.factor(nombre_es))
+
+# Load dataset and manipulate addresses.
+df = read.csv("data/original/190302_viviendas-turisticas-comunidad-valenciana_valencia.csv") %>%
+  separate(Address, c("nombre_es_raw", "num"), extra = "merge",
+           sep = ", nº", remove = FALSE) %>%
+  separate(num, c("num", "puerta"), extra = "merge", sep = ", ", remove = FALSE) %>%
+  mutate(nombre_es_raw = as.factor(toupper(nombre_es_raw)))
+
+
+
+# Basic Record linkage.
+record.linkage.names = function(names,
+                                officialnames = calles_valencia$nombre_es) {
+  # Matches a street's name with the most similar official one.
+  for (i in names) {
+    print(i)
+    inferred.name = as.character(
+      officialnames[amatch(i, toupper(officialnames), maxDist = 5)])
+    if (!is.na(inferred.name)) {
+      names[names == i] = inferred.name
+    }
+  }
+  return(names)
+}
+
+
+# df$nombre_es = record.linkage.names(df$nombre_es_raw)
+
+df = df %>%
+  mutate(nombre_es = record.linkage.names(nombre_es_raw)) %>%
+  left_join(calles_valencia)
 
 # Geocoding with Photon ---------------------------------------------------
 
-# require(devtools)  
-# devtools::install_github(repo = 'rCarto/photon')  
-library(photon)
-library(dplyr)
 
-df = read.csv("https://code.montera34.com/airbnb/valencia/raw/master/data/original/190302_viviendas-turisticas-comunidad-valenciana_valencia.csv") %>% 
-  mutate(Address = gsub("C ", "calle de ", Address)) %>% 
-  mutate(Address2 = paste(Address, Municipio, sep = ", "))
-  
-geocoded.df = photon::geocode(head(df$Addresses2), limit = 1,
+df2 = df %>%
+  select(Signatura, Municipio, tipovia_ca, nomoficial, num) %>%
+  mutate(full_address_ca = paste(tipovia_ca, nomoficial, num, Municipio,
+                                 sep = ", "))
+
+geocoded.df = photon::geocode(head(df2$full_address_ca), limit = 1,
                               # lang = "es",
                               key = "highway",
                               locbias = c(-0.3766, 39.4665))
 
 
 # Combine geocoded dataframe with original one.
-df.combined = geocoded.df %>% 
-  select(location, lon, lat) %>% 
+df.combined = geocoded.df %>%
+  select(location, lon, lat) %>%
   right_join(df, by = c("location" = "Address2"))
-  
-write.csv(df.combined, file = "../data/output/filename.csv")
\ No newline at end of file
+
+write.csv(df.combined, file = "data/output/filename.csv")

valencia.Rproj

+Version: 1.0
+
+RestoreWorkspace: Default
+SaveWorkspace: Default
+AlwaysSaveHistory: Default
+
+EnableCodeIndexing: Yes
+UseSpacesForTab: Yes
+NumSpacesForTab: 2
+Encoding: UTF-8
+
+RnwWeave: Sweave
+LaTeX: pdfLaTeX
+
+AutoAppendNewline: Yes
+StripTrailingWhitespace: Yes