Potential features
| Variable name | Description | Justification | How to fill missing data |
|---|---|---|---|
var_name |
DESCRIPTION | THOUGHTS ON WHY IT MIGHT BE USEFUL | HOW SHOULD MISSING DATA BE FILLED IN? |
beach_direction |
North Beach or South a Beach? Manual Clustering of beach variable | Navy pier appears to be a useful geographic split for clustering beaches with similar amounts of high E. coli levels | NA |
beach_cluster |
Like beach direction, but with more levels. Can be determined by inspection or through a clustering technique. | Allows a model to more fine-tune estimates within clusters. | NA |
change_in_previous_reading_Dn |
Prior day minus its day n prior. (for n = 1, 2, 3) |
If bird poop is accumulating, a large increase in one reading may anti-correlate with future readings. | TODO |
days_since_last_substantial_rain |
Number of days since the last time the beach was cleaned. | Again on the topic of accumulating bird poop. | TODO |
inches_rainfall_past_x |
Number of inches in the past x days (for x = .5, 1, 2, 3, 4, 7) |
Same | TODO |
| historical daily weather summaries | ... | ... | ... |
| hourly weather from recent history (5am day before through 4am day of?) | ... | ... | ... |
| Cross-beach info (e.g. the E. coli readings of each of the groups, perhaps?) | ... | ... | ... |
| sunrise time | ... | ... | ... |
Looking ahead to more variables that are available only from 2015 onwards:
| Variable name | Description | Justification | How to fill missing data |
|---|---|---|---|
| water sensor data (turbidity, etc.) | ... | ... | ... |
| USGS model prediction | ... | ... | ... |
