#Detecting Fraud, Corruption and Collusion in World Bank Contracts - Data Science for Social Good Summer 2015 Fellowship at the University of Chicago

##PROJECT OVERVIEW This project focused on detecting fraud, collusion and corruption in the World Bank contracts. Corruption is the biggest impediment to economic growth in 60+ developing countries and drains about $900+ billion from the developing world every year.

The World Bank Group loans money to developing countries so that they can undertake infrastructure and development projects. During this money loaning process, countries post RFPs (requests for proposals or bids) and contractors submit bids. Winning companies which are awarded a contract subsequently begin work and begin to bill for that work. Occasionally, during the bidding and billing process, companies may engage in corrupt behavior. The World Bank Group currently relies on complaints and the domain expertise of contract investigators to identify wrong-doing.

The DSSG team built data science models using machine learning techniques to:

• Prioritize complaints by using historical records of past World Bank - contracts and investigation results.
• Pro-actively identify contracts to investigate in addition to relying on complaints.

##SETUP INSTRUCTIONS To Install from Source Code contained in this directory - please follow the instructions below.

• Git clone https://github.com/eredmiles/Fraud-Corruption-Detection-Data-Science-Pipeline-DSSG2015.git

You will run full_pipeline.sh to do EVERYTHING!

• Install Anaconda python (see instructions here NOTE: Install Python 2.7 NOT Python 3.X

• Setup PostgreSQL database:

  1. sudo apt-get install postgresql postgresql-contrib
  2. Install according to these instructions

• Install pandas - conda install pandas

• Install seaborn - conda install seaborn

• Install csvkit - conda install csvkit

• Install psycopg2 - conda install psycopg2

###Required Files

• The investigations file must be saved as a .csv file. Further, it must be named investigations.csv
• Canonical entity resolution file already included in pipeline_data (canonicalFileV2.csv)
• Download the zip file from here: http://data.worldbank.org/indicator/PA.NUS.PPP, save the ppp file within as ppp.csv in pipeline_data (or your DATA_STORAGE variable location, see below). Do the same for the zip file from here: http://data.worldbank.org/indicator/PA.NUS.FCRF, save the fcrf file as fcrf.csv
  • A 2015 version of these files is present in pipeline_data as ppp.csv and fcrf.csv

###Modify full_pipeline.sh as follows:

• The LOCALPATH variable should be the path to the WorldBank2015 directory that you cloned from git (e.g. /dssg/Fraud-Corruption-Detection-Data-Science-Pipeline-DSSG2015)
• The DATA_STORAGE variable should be the path where your data files AND THE INVESTIGATION FILE will live. there is already a directory called pipeline data in the github repo that contains the entity resolution canonical. You should use this directory (e.g. /dssg/Fraud-Corruption-Detection-Data-Science-Pipeline-DSSG2015/pipeline_data)

NOTE: This is also where all output files will be stored.

• The CURRENCY_FILE_PPP variable should contain the full path where you want ppp.csv (a currency conversion file) to live.

• THE CURRENCY_FILE_FCRF variable should contain the full path where you want fcrf.csv (a currency conversion file) to live.

• If you are no longer using the local host database:

  • you must change the database connection in sql.py (Line 54), supplier_feature_gen (line 42) Example: host="localhost",user="dssg",password=password,dbname="world_bank" copy example_config to config (e.g. cp example_config config), modify config to have the password for the database (e.g. the system password for your user, in our example the user password for dssg)
  • you must also modify model_pipeline_script.py (line 86) and supplier_feature_gen.py (line 217): create_engine(r'postgresql://[USER_NAME]:'+password+'localhost/DATABASE'. e.g. create_engine(r'postgresql://dssg:'+password+'localhost/world_bank'
  • you must create a config file in the directory from which you will run the script.

##Authors

Emily Grace	Ankit Rai	Elissa Redmiles
Princeton University	University of Illinois Urbana-Champaign	University of Maryland
Ph.D. Candidate in Physics	Ph.D. Student in Informatics	Ph.D. Student in Computer Science
Contact: [email protected]	Contact: [email protected]	Contact: [email protected]

##Licensing All contents are covered under an MIT license. See the LICENSE file in this directory for further information.