In rockclimber112358/romeHousePrices: Package to scrape and analyze Rome housing prices

Load the libraries and the Mio Affitto dataset:

suppressMessages({
library(data.table)
library(ggplot2)
library(scales)
library(glmnet)
library(sqldf)
library(reshape2)
library(romeHousePrices)
})

assignDirectory()

load(paste0(savingDir, "/detail_Mio_2015.10.05.06.28.32_cleaned.RData"))

## When filtering to top zones, quartieres, etc. how many should be kept?
maxToPlot = 10

Set some options:

## Keep data.table from printing all the time
assign("depthtrigger", 100, data.table:::.global)

Prezzo vs. Superficie

d[, roundSuper := round(superficie/25) * 25]
print(ggplot(d, aes(x = prezzo, fill = factor(roundSuper))) +
    geom_bar(position = "fill") +
    xlim(c(0,3001)) + labs(y = "", fill = "superficio") +
    scale_y_continuous(labels = percent))

Exploring the model

Locali

Not surprisingly, price increases with number of rooms. However, it's interesting to see that the increase seems to be faster than linear: prices really jump after 4 or 5 rooms. But, that may just mean that we've moved into a group of really nice houses, whereas the 1-4 locali houses are a mix of fancy and not so fancy.

ggplot(d, aes(x = locali, y = prezzo)) +
    geom_point() + geom_smooth()

ggplot(d, aes(x = locali, y = prezzo, group = locali)) +
    geom_boxplot()

Superficie

ggplot(d, aes(x = superficie, y = prezzo)) +
    geom_point() + geom_smooth()

ggplot(d, aes(x = superficie, y = prezzo / superficie)) +
    geom_point() + geom_smooth()

Bagni

ggplot(d, aes(x = bagni, y = prezzo)) +
    geom_point()

ggplot(d, aes(x = bagni, y = prezzo, group = bagni)) +
    geom_boxplot()

Piano

Zona

d[, zonaCnt := .N, by = zona]
cutoff = sort(unique(d$zonaCnt), decreasing = TRUE)[maxToPlot]
ggplot(d[zonaCnt >= cutoff, ], aes(x = substr(zona, 1, 30), y = prezzo,
                                   group = zona)) +
    geom_boxplot() + theme(axis.text.x = element_text(angle = 90))
d[, zonaCnt := NULL]

Quartiere

d[, quartiereCnt := .N, by = quartiere]
cutoff = sort(unique(d$quartiereCnt), decreasing = TRUE)[maxToPlot]
ggplot(d[quartiereCnt >= cutoff, ], aes(x = quartiere, y = prezzo, group = quartiere)) +
    geom_boxplot() + theme(axis.text.x = element_text(angle = 90))
d[, quartiereCnt := NULL]

CAP

d[, capCnt := .N, by = CAP]
cutoff = sort(unique(d$capCnt), decreasing = TRUE)[maxToPlot]
ggplot(d[capCnt >= cutoff, ], aes(x = CAP, y = prezzo, group = CAP)) +
    geom_boxplot() + theme(axis.text.x = element_text(angle = 90))
d[, capCnt := NULL]

Results of Model

• Prediction
• Residuals (look for trends)
• Best and worst deals

Time series

• Average price over time: ggplot with many curves
• Average price over time: spatially with many images to capture time component
• One spatial plot indicating the percent change in house prices over some time period by region.

Partition data by - CAP - Zona - Quartiere

Simple spatial plots

• Price density
• Number of houses in region
• Number and/or quality of services in region
• Covariates (average superficie, average numeri di bagni/locali, ...)

rockclimber112358/romeHousePrices documentation built on May 27, 2019, 12:15 p.m.