This repository includes all the data pre-processing scripts for the project PRISMS.

The script is generating a grid map over the target region in Postgres.
Table GRID: [gid, centroid, lon, lat, geom, lon_proj, lat_proj]

Input parameters:
- the bounding box over the target area
- the EPSG of the target area (unit should be "metre")
- the resolution of the grid
- the grid table object

The script is computing the values of various geographic features within each cell from OpenStreetMap.
Table GEO_FEATURE: [gid, feature_type, geo_feature, value, measurement]

Input parameters:
- the bounding box over the target area (used for cropping)
- the OpenStreetMap table objects and corresponding geo features (e.g., landuses, roads)
- the grid table object
- the geo feature table object

The script is constructing a geo vector from the geo features.
Table GEO_VECTOR: [gid, data] # data is a list
Table GEO_NAME: [name, geo_feature, feature_type]

Input parameters:
- the grid table object
- the geo feature table object
- the geo vector table object
- the geo name table object

The script is mapping the grid to a matrix (re-indexing).
An example of the output matrix:
[[6917, 6918, 6919, ..., 6990, 6991, 6992],
 [6841, 6842, 6843, ..., 6914, 6915, 6916],
 [6765, 6766, 6767, ..., 6838, 6839, 6840],
 ... ...,
 [153, 154, 155, ..., 226, 227, 228],
 [77, 78, 79, ..., 150, 151, 152],
 [1, 2, 3, ..., 74, 75, 76]])

Input parameters:
- the grid table object
- the output filename  # the output would be .npz file

The script is randomly generating training, validation, and testing locations with evenly-spatial distribution.

Input parameters:
- the given locations
- the number of pieces dividing the space or the number of clusterst  # extracting locations from each cluster to ensure even distribution

The script is interpolate the features (meteorological) across the time using linear interpolation.
Table INTERPOLATION: [gid, timestamp, data]

Input parameters:
- the old meteorological table object
- the target meteorological table object  # having the same spatial resolution as the old one

The script is interpolate the features (meteorological) across the space using cubic interpolation.
Table INTERPOLATION: [gid, timestamp, data]

Input parameters:
- the old grid table object
- the old meteorological table object
- the target grid table object
- the target meteorological table object  # having a finer spatial resolution than the old one

The script is generating the training data including label matrix and feature matrix.
The output file contains "label_mat", "feature_mat", "feature_distribution", "geo_name", "pm_grids", "grids".

Input parameters:
- the air quality table object
- the meteorological table object
- the geo vector table object
- the geo name table object
- the grid table object & the mapping matrix file
- the time range
- the output filename  # the output would be .npz file