Ultra Sound Garbage Bin Sensor Reading Records

The directory is structured in the following way:

The file "ids.csv" is a one-column table that contains the IDs of all the containers present in the dataset. Use this file as a helper to iterate over the dictionary.

The file "INFO.csv" contains relevant data about the containers, namely "Capacity," "WasteType," "Latitude," and "Longitude."

Then, for each container, there are four files with the following naming convention:

Container_ID_recs_UnCorrected_with_metrics.csv

Container_ID_fill_UnCorrected_with_metrics.csv

Container_ID_recs_Corrected_with_metrics.csv

Container_ID_fill_Corrected_with_metrics.csv

The RAW values extracted from the sensors are in the UnCorrected files. The FILL levels are extracted by the sensors, and DATES (whether from collections or sensor readings) are provided by the management system. The sensor timestamps are automatic, and collection times are manually entered by service providers. The rest of the information consists of calculations made by us for reproducibility. An atempt to correct some data was preform, these are present in the corrected files.

The fill files contain information about the Fill level of the container. Additional columns include the Max and Min monotonic approximations between each collection, along with their mean (middle point) value. The "REC" column is a mask set to true for the first fill value measured after a collection is made. The "Cidx" column is a unique identifier for each fill value between collections, which allows us to use groupby on each event between collections.

The rec files contain information about the Collections made for that container. Additional columns include the End_Pointer, which is the row index for when that collection is made in the corresponding fill file, along with AVG_DIST, which is 100 minus the average difference between the Max and Min approximations until the next collection, and 100 times the Spearman coefficient for the fill values until the next collection.