Time-based

Loading Data

airbnb <- read_delim("./airbnb_austin.csv", delim = ",")

## Rows: 15244 Columns: 18
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr   (3): name, host_name, room_type
## dbl  (12): id, host_id, neighbourhood, latitude, longitude, price, minimum_n...
## lgl   (2): neighbourhood_group, license
## date  (1): last_review
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Preparing Time Data

airbnb_ <- airbnb |>
  select(last_review, price, room_type) |>
  distinct()

#airbnb_

Response Variable

airbnb_ts <- airbnb_ |>
  filter(!is.na(last_review)) |>
  mutate(date = as.Date(last_review)) |>
  group_by(date) |>
  summarise(avg_price = mean(price, na.rm = TRUE)) |>
  as_tsibble(index = date) |>
  fill_gaps() |>
  mutate(avg_price = na.approx(avg_price, na.rm = FALSE))

airbnb_xts <- xts(x = airbnb_ts$avg_price, 
                  order.by = airbnb_ts$date)

airbnb_xts <- setNames(airbnb_xts, "airbnb")

airbnb_ts |>
ggplot() +
  geom_line(mapping = aes(x=date, y=avg_price)) +
  theme_hc()

## Warning: Removed 360 rows containing missing values or values outside the scale range
## (`geom_line()`).

• There’s a sudden sharp spike around 2022 that needs to be investigated further, which might be due an anormalie or event.

airbnb_xts |>
  rollapply(width = 30, \(x) mean(x, na.rm = TRUE), fill = FALSE) |>
  ggplot(mapping = aes(x = Index, y = airbnb)) +
  geom_line() +
  labs(title = "Airbnb Average Price Trends Over Time",
       subtitle = "30-Day Rolling Average of Listing Prices Based on Last Review Date") +
  theme_hc()

airbnb_xts |>
  rollapply(width = 30, \(x) mean(x, na.rm = TRUE), fill = NA) |>
  fortify.zoo() |>
  ggplot(mapping = aes(x = Index, y = airbnb)) +
  geom_line() +
  scale_x_date(date_labels = "%b %Y", date_breaks = "4 months") +  # Formats axis as "Jan 2024", etc.
  labs(title = "Airbnb Average Price Trends Over Time",
       subtitle = "30-Day Rolling Average of Listing Prices Based on Last Review Date",
       x = "Date", y = "Average Price") +
  theme_hc() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

## Warning: Removed 360 rows containing missing values or values outside the scale range
## (`geom_line()`).

• Prices peak during summer months (June-August)

• Dip in prices visible December - April

Trend Analysis

airbnb_ts <- airbnb_ts |> 
  arrange(date) |>
  rename(date = date)

model_trend <- airbnb_ts |>
  model(trend_model = TSLM(avg_price ~ trend()))

airbnb_dc <- as.ts(airbnb_ts) |>
  decompose()   

plot(airbnb_dc)

• Price generally trends upwards, possibly due to increase demand or inflation

• The summer period shows massive price surges which likely due to holidays

• The residual spikes suggest the model doesn’t fully explain all price dynamics, might need further investigation.

Seasonal Analysis

airbnb_ts |>
  ACF(avg_price) |>
  autoplot() +
  labs(title = "ACF: Autocorrelation of Average Price")

airbnb_ts |>
  PACF(avg_price) |>
  autoplot() +
  labs(title = "PACF: Partial Autocorrelation of Average Price")

• The autocorrelations remain strong and statistically significant for many lags, indicating that past Airbnb prices have a significant influence on future prices.

• Hosts adjust prices based on past performance.

• The PACF has a significant spike at lag1 and much weaker spikes beyond that, which might suggest the current price is mostly influenced by the immediately preceding value, with diminishing effect from earlier lags.

Summary

• Strong seasonal patterns, with the summer period experiencing the highest prices.

• Yesterday’s or last week’s price is the biggest influence on today’s price

• Hosts might adjust pricing seasonally, which can be influenced by events and seasons, e.g Christmas or summer.