Space Apps Challenge 2018

#OUR TEAM

#Why Mosquitalert?

We hope the citizen scientist idea could let people care more about their living environment and help public health official to collect data at the ground level by playing our mobile game.

#Problem We Solve

Collect useful data used to be complicated and difficult for people don’t have measure equipment. With Mosquitalert you can easily collect data from mobile phone to help public health official and became a citizen scientist.

#Our Goal

Our goal is dengue fever and Zika virus will not pose a threat to the world in the future.

#GitHub

• iOS App: https://github.com/PlanetBroccoli/BroccoliEarth
• API: https://github.com/PlanetBroccoli/laravel_api
• Data Analysis: https://github.com/PlanetBroccoli/NASA

#How it works?

• Mobile app:

  • Are you in the high-risk area? Show analyzed data on the view
    

• Wanna see the cute AR mosquito? Show how many reports nearby with mosquito level
  

• Let’s level up! See the mission list to become a citizen scientist step by step

• Prevent! Take pictures to tell us where might cause mosquito breeding

• Got bites? Report it!
  

• Tell me more! More information
  

• Website:

  web

  • See the prediction of vector mosquito area

• See the location and picture of user reports

#Technologies We Use

• Data Analysis : Python, R
• Backend : PHP, Laravel, Flask
• Web : React, ES6, SSR, Html5, CSS3
• Mobile : Swift, Alamofire, ARKit,
• Database : MySQL
• Cloud : AWS EC2, AWS S3

#How we made it:

Data Analysis Methodology

Based on statistical analysis, five dengue fever cases per week in a district represents the 80th percentile of our dataset and is the threshold which we base our “threat level” on (i.e. the “threat level” is high if we predict there to be at least five dengue fever cases).

We used a logistic regression model to calculate the likelihood of five or more dengue cases occurring in a district within a week.

Choice of Data

1. Housing and population data (https://census.dgbas.gov.tw/PHC2010/english/rehome.htm)
   Data Relevance: Older housing estates and buildings are locations which see higher instances of mosquito breeding, hence increasing the likelihood of mosquito-related illnesses such as dengue fever. High population density increases the spread of dengue fever.
2. Meteorological data (http://e-service.cwb.gov.tw/HistoryDataQuery/)
   Data Relevance: High rainfall, low wind speeds and low pressure likely provide an ideal environment for mosquitoes to thrive and thus we expect to see higher instances of mosquito-related illnesses in such weather conditions.
3. Industry and agricultural data (https://eng.stat.gov.tw/ct.asp?xItem=37575&ctNode=1634&mp=5)
   Data Relevance: Agriculture, farming and forestry operations affect the environment and may create opportunities for mosquitoes to thrive (examples include stagnant water in farms).
4. Geographical features data / other surrounding facilities (https://opendata.epa.gov.tw/)
   Data Relevance: Geographical features and surrounding facilities may contribute to growth of mosquitoes (examples include hot springs, where stagnant water builds up easily or landfills, which may breed pests more easily).

Training Parameters

The parameters we used are as follows:

• Housing and population data
  o District population
  o Surface area
  o Unoccupied housing
  o Occupied housing
  o Housing building height (number of floors)
  o Housing building age (years since construction)
  o Number of occupants
• Meteorological data
  o Wind speed
  o Maximum gust
  o Temperature (of the surroundings)
  o Maximum temperature
  o Minimum temperature
  o Humidity
  o Atmospheric pressure
  o Precipitation
• Industry and agricultural data
  o Total cultivated area (farmland / agriculture)
• Geographical features data / other surrounding facilities
  o Number of (notable) hot spring locations
  o Number of landfill locations
  o Forest land area
• Total number of parameters: 19
• Total number of data points: 585 (can be increased greatly with more time)

JSON format of the dataset can be provided on request.

Some interesting findings

• Landfill is a very significant predictor of dengue fever occurrence (p-value = 0.000353).
  o Landfills are usually located in the suburbs where mosquito control mechanisms may be lacking
  o Rubbish and poor landfill management can rapidly increase mosquito spawn
• Population is (obviously) a very significant predictor of dengue fever (p-value = 0.0000182), but land surface area and population density are not significant predictors.
  o The influence of land surface area may have been reduced by the average number of people living in each residential unit
• Locations with residential units between 0 and 10 years of age and 10 and 20 years of age were also significant predictors of dengue fever occurrence (p-values 0.003296 and 0.000157 respectively)
  o Newer buildings may indicate wealth and ability to control mosquito populations.
  o Furthermore, newer buildings may have fewer mosquito breeding hazards
• Locations with high average number of people per residential unit are also significant predictors of dengue fever occurrence (p-value = 0.003397)
  o This could be due to high proximity of contact and is a better indicator than population density.

Application interesting finding :

• Using user friendly interface and interaction to evoke the awareness of the population
• Getting the general ideas of AR, 3D model and building game
• Using the third-party to combine social media (Facebook login/ sharing...etc) to disease prediction or protection