Project 4a: Predicting the Spread of West Nile Virus

Problem Statement

Team 39 SIR of the Disease And Treatment Agency, division of Societal Cures In Epidemiology and New Creative Engineering (DATA-SCIENCE) is to effectively plan the deployment of pesticides in order to mitigate the spread of the West Nile Virus in Chicago City.

This will be done by analysing the data given by the Department of Public Health to produce the following deliverables:

1. A predictive model to facilitate informed decision making by the city of Chicago when it decides where to spray the pesticides.

2. Cost-Benefit Analysis of the annual cost projections for various levels of pesticide coverage (cost) and the effect of these various levels of pesticide coverage (benefit).

Background

West Nile virus (WNV) is the leading cause of mosquito-borne disease in the continental United States.

It is most commonly spread to people by the bite of an infected mosquito. Cases of WNV occur during mosquito season, which starts in the summer and continues through fall.

Most people infected with WNV do not feel sick:

• About 1 in 5 people who are infected develop a fever and other symptoms
• About 1 out of 150 infected people develop a serious, sometimes fatal, illness.

Source: https://www.cdc.gov/westnile/index.html

Datasets

There are 4 datasets sponsored by the CDC: train, test, weather and spray.

Data Dictionary

train_df

Period: 2007, 2009, 2011, and 2013

Feature	Type	Description
Date	object	Date that the WNV test is performed
Address	object	Approximate address of the location of trap, this is used to send to the GeoCoder
Species	object	The species of mosquitos
Block	int64	Block number of address
Street	object	Street name
Trap	object	Id of the trap
AddressNumberAndStreet	object	Approximate address returned from GeoCoder
Latitude	float64	Latitude returned from GeoCoder
Longitude	float64	Longitude returned from GeoCoder
AddressAccuracy	int64	Accuracy returned from GeoCoder
NumMosquitos	int64	Number of mosquitoes caught in this trap
WnvPresent	int64	Whether West Nile Virus was present in these mosquitos. 1 means WNV is present, and 0 means not present.

test_df

Period: 2008, 2010, 2012, and 2014

Feature	Type	Description
Id	int64	The id of the record
Date	object	Date that the WNV test is performed
Address	object	Approximate address of the location of trap, this is used to send to the GeoCoder
Species	object	The species of mosquitos
Block	int64	Block number of address
Street	object	Street name
Trap	object	Id of the trap
AddressNumberAndStreet	object	Approximate address returned from GeoCoder
Latitude	float64	Latitude returned from GeoCoder
Longitude	float64	Longitude returned from GeoCoder
AddressAccuracy	int64	Accuracy returned from GeoCoder

weather_df

Period: 2007, 2008, 2009, 2010, 2011, 2012, 2013, and 2014

Feature	Type	Description
Date	object	Date of record
Station	int64	Station number, either 1 or 2
Tmax	int64	Maximum temperature in Degrees Fahrenheit
Tmin	int64	Minimum temperature in Degrees Fahrenheit
Tavg	object	Average temperature in Degrees Fahrenheit
Depart	object	Temperature departure from normal in Degrees Fahrenheit
DewPoint	int64	Average Dew Point in Degrees Fahrenheit
WetBulb	object	Average Wet Bulb in Degrees Fahrenheit
Heat	object	Absolute temperature difference of Tavg from base temperature of 65 Degrees Fahrenheit if Tavg < 65
Cool	object	Absolute temperature difference of Tavg from base temperature of 65 Degrees Fahrenheit if Tavg > 65
Sunrise	object	Time of Sunrise (Calculated, not observed)
Sunset	object	Time of Sunset (Calculated, not observed)
CodeSum	object	Weather Phenomena, refer to CodeSum Legend below
Depth	object	Snow / ice in inches
Water1	object	Water equivalent of Depth
SnowFall	object	Snowfall in inches and tenths
PrecipTotal	object	Rainfall and melted snow in inches and hundredths
StnPressure	object	Average station pressure in inches of HG
SeaLevel	object	Average sea level pressure in inches of HG
ResultSpeed	float64	Resultant wind speed in miles per hour
ResultDir	int64	Resultant wind direction in Degrees
AvgSpeed	object	Average wind speed in miles per hour

spray_df

Period: 2011, and 2013

Feature	Type	Description
Date	object	Date of the spray
Time	object	Time of the spray
Latitude	float64	Latitude returned from GeoCoder
Longitude	float64	Longitude returned from GeoCoder

Recommendations

WHEN

to spray, based on the total number of mosquitoes trapped monthly.

Monitoring: Traps Continue from May through Oct

Action: Spray When traps hit 14% WNV-positive

WHERE

to spray, based on the top 10 hot spots.

1. ORD Terminal 5, O'Hare International Airport
2. South Doty Avenue
3. 4100 North Oak Park Avenue
4. South Stony Island Avenue
5. 4600 Milwaukee Avenue
6. 8200 South Kostner Avenue
7. 2400 East 105th Street
8. 3600 North Pittsburgh Avenue
9. O’Hare Court, Bensenville
10. 7000 North Moselle Avenue

Future Research

• Analyse effect of birds on WNV infection

  • Birds are amplifying hosts (Environmental Research and Public Health, 2020)

• Analyse the severity of WNV cases

  • Look at total no. of cases instead of binary outcomes