Kaggle Paris Meetup

• Meetup #11
• 3 Proposals :

1. Data Science Bowl 2017​(image),large dataset​
2. The Nature Conservancy Fisheries Monitoring (NCFM)

3. DengAI: Predicting Disease Spread

Data Science Bowl 2017

• Image Classification  (voxel) 

• $1,000,000 · 1,116 teams · 2 months to go (2 months to go until merger deadline)

The Nature Conservancy Fisheries Monitoring (NCFM)

• Image classification 
• The Nature Conservancy Fisheries Monitoring
• Repo with InceptionV3 utilization from keras.
• https://github.com/pengpaiSH/Kaggle_NCFM

DengAI: Predicting Disease Spread

Open until dec 17. For the glory.

The goal is to predict the total cases for two cities :

• Iquitos (Peru)
• San Juan (Puerto Rico)

​

The training set mainly consist of quantitative data :

• 24 columns
• 1456 rows (936 for San Juan, 520 for Iquitos)
• 20 years of data ranging from 1990 to 2010
• 1 "city" column
• 3 dates columns (year, week of year, week start date)
• 4 columns about Satellite Vegetation (in pixel, N/S/E/W) 

DengAI: Predicting Disease Spread

• 6 columns about precipitation and humidity
• 10 columns about temperature

next slide : General information about the problem at hand

DengAI: Predicting Disease Spread

General information about the problem at hand :

• Deng is a disease caused by a virus (Flavivirus)
• This virus is spread by  only one specific type of mosquito (Andes aegypti)
• This specific mosquito  :
  • has to be a female in order to spread the virus
  • has to bites someone with the deng fever in order to cary the virus
  • has to incubate for 8 days (avg)
  • then has to bite someone (often operates when it's daylight)
  • does not travel more than 100 meters from its home
  • can live two to three weeks
  • Flavivirus is a family of virus where we find Zika, Yellow Fever, Aedes is also a known vector for Chikungunya

DengAI: Predicting Disease Spread

Accurately predict the number of next cases can contribute to the set of existing  responses for a global situation