# Version 5 Description ## TLDR - The directory **1-1-16_5-31-17\_Weather** contains 1663 files (one for each of the 1663 stations in Japan) - As its name implies, the data is from the same date window as the competition's **date\_info** file - If a station file's name ends with four underscores and a date, the date indicates when the station was terminated - Please, read on... ## Context This dataset contains the dataset from the Recruit Restaurant Visitor Forecasting competition (active from 11-28-17 to 2-6-18).
The focus is on using data about reservations made at various restaurants throughout Japan, along with restaurant location and genre information to predict the actual number of visitors a restaurant will have on a given day. This dataset augments the above with the addition of information about the weather at various locations in Japan over time to produce an exciting, multi-faceted dataset that deals with time, geography, weather, and delicious food. Thank you for your interest in this dataset! Please let me know if you have any suggestions, questions or problems! ## Content The core of the dataset comprises the files in the following directory: 1-1-16_5-31-17_Weather (1663 .csv files): This directory contains translated weather data for the time period denoted by the directory’s name (from 1-1-16 through 5-31-17).
Each .csv file in this directory is of shape (517, 15), and is named according to the id values in the below **weather\_stations** file.
There are a few reasons why there may seem to be a lot of null values: - The primary reason is that different types of stations/sensors are used, and some just don't capture as much data as others - Questionable data is sometimes removed by the Agency - If the station was terminated, its values are null These are the features for all translated weather files (I won’t hazard a description for the features that aren’t already self-explanatory because I’m no meteorologist, and I’d hate for you to get that impression): - calendar_date \- the observation date, formatted thusly "yyyy-mm-dd" - avg_temperature - high_temperature - low_temperature - precipitation - hours_sunlight - solar_radiation - deepest_snowfall - total_snowfall - avg_wind_speed - avg_vapor_pressure - avg_local_pressure - avg_humidity - avg_sea_pressure - cloud_cover --- This dataset adds the following .csv files regarding weather stations, and their relations to the competition data: weather\_stations.csv (1663, 8): This file contains the location and termination dates for 1,663 weather stations in Japan. - id \- the join of a station’s prefecture, first_name, and second_name, with "\_\_" (double underscores) - Note: If date_terminated is not null, id will end with four underscores and the date_terminated - prefecture \- the prefecture in which this station is located \(see note 1\) - first_name \- the first name given to specify a location \(see note 2\) - second_name \- the second name given to specify a location \(see note 2\) - latitude \- latitude of the station, converted from degrees, minutes, seconds to decimal degrees for consistency - longitude \- longitude of the station, converted from degrees, minutes, seconds to decimal degrees for consistency - altitude \- altitude of the station - date_terminated \- If the station was terminated, the date of its termination \(formatted thusly "yyyy-mm-dd"\) else null nearby\_active\_stations.csv \(62, 8\): This file is a subset of weather\_stations.csv (above) selected via the following criteria:
    1\) the station was not terminated, and
    2\) the station was the closest station to at least one store in **air\_store\_info** or **hpg\_store\_info**
As you can see, there is a lot of overlap here, because while the weather stations seem to generally be scattered throughout Japan, the store locations tend to be clustered around several areas. Column names and descriptions are identical to those of **weather\_stations**. feature\_manifest.csv (1663, 15): This file contains information about each station's "coverage" of each weather feature.
Values of 0.0 for any of the below features except id mean that station collected no data on that feature.
Values of 1.0 for any of the below features except id mean that station collected data on that feature for every day. - id \- the id of this weather station - avg_temperature \- ratio of non-null values for this feature at this station - high_temperature \- ratio of non-null values for this feature at this station - low_temperature \- ratio of non-null values for this feature at this station - precipitation \- ratio of non-null values for this feature at this station - hours_sunlight \- ratio of non-null values for this feature at this station - solar_radiation \- ratio of non-null values for this feature at this station - deepest_snowfall \- ratio of non-null values for this feature at this station - total_snowfall \- ratio of non-null values for this feature at this station - avg_wind_speed \- ratio of non-null values for this feature at this station - avg_vapor_pressure \- ratio of non-null values for this feature at this station - avg_local_pressure \- ratio of non-null values for this feature at this station - avg_humidity \- ratio of non-null values for this feature at this station - avg_sea_pressure \- ratio of non-null values for this feature at this station - cloud_cover \- ratio of non-null values for this feature at this station air\_station\_distances.csv (1663, 111): This file contains the Vincenty distance from every weather station to every unique latitude/longitude pair in the air system. - station_id \- the id of this weather station - station_latitude \- station latitude (in decimal degrees) - station_longitude \- station longitude (in decimal degrees) - <