Problem Statement

In today’s world, moving out and moving in is synonymous with the profession we undertake. In this global Organization our work and personal demands take us to many places and eventually sooner then we expect we are faced with the challenge to look for a place to live, in a new locality. In today’s world and in big cities like Chicago, invariably we find ourselves to make an important decision regarding location to buy a second home , get rented place , move out to a new neighbourhood or move near to location of work. These problems are real and more so important, as decisions like buying a house or renting a place affect us in more ways than we can think of. We from Accenture are proposing an idea in MIT Alliance challenge 2014 that will help us make this predictive decision on one of the most important aspect of our lives. We are planning to use Big Data and mobile technologies along with other next gen technology to help us make important decisions of life.

Solution Offering

Our solution offering in this MIT Alliance challenge 2014 is to take Chicago crime data for last 10 years and build a predictive model. This model will work on Hadoop Big data Implementation. The first part of this solution proposed will run a predictive model on the huge data set (last 10 years and counting) and generate analytics on the locality which are more susceptible to crime. The Second part of this solution offering will have mobile applications in android and iOS which will connect to analyzed data generated on previous step and give a real time suggestions , breakups and graphs to help make decision whether to move to the searched locality. Over all the solution we have created will require user to open mobile application and search a specific zip code where he/she intends to go. The mobile application will talk to a noSQL database using web services and will get details of the locality along with predictive recommendation if this area is really safe. These in depth information will help the user to make a rational decision to move to the new locality. Moreover at the background every weekend we schedule a new Hadoop job which when executed will analyze the new crime data from the city of Chicago and update the noSQL database with more recent analytics and findings.

Technology Stack

The solution offering has the following technology stack as part of the solution offering.

Solution Architecture

This section explains the salient features of the solution offering.

• Data is loaded every weekly (or as intended) into S3 buckets in AWS.

• Custom MapReduce Analytics job reads the data from S3 and performs predictive analytics.

• The analytical data is loaded from HDFS into Data Store using SQOOP process.

• Users use mobile application/tablets to search locality by zip code.

• The mobile applications use REST web services to access the analytical recommendations from the database.

• Users get a locality recommendation based on the huge historical crime data available with the Chicago Police.

The MIT Alliance 2014 Data Science Challenge

This section demonstrates the working of the proof of concept for the solution offering of The Accenture and MIT Alliance 2014 Data Science Challenge discussed above. This section is divided into 2 parts – one being the Big data Hadoop analytics part and another being the client facing mobile application part.

Part 1 – Analytics Part: Big Data

The POC is built on 3 node small cluster on Cloudera Hadoop distribution. This cluster is hosted on AWS. (Due to the cost of operating the cluster , we would request to contact us for a live demo as we normally keep the cluster turned off when not in use). NOTE:: Please contact us for a live demo on the analytics part , the visualization part in mobile is available for use. This section provides a screenshot walkthrough of the Analytics process. Step 1 Data is loaded into S3 using s3cmd or S3 browser. This step is executed once per week (or on demand).

Step 2 Custom MR job is executed as oozie workflow to get the data from S3 and process the data using custom predictive and ranking algorithm to generate analyzed data into HDFS.

Step 3 Sqoop job then runs on the output from Map Reduce custom code and uploads the analytical data into MySQL table.

Part 2 – Visualization using mobile apps

This section is functional and testable. This part describes how the analytical data generated from the previous steps can help to make a real time decision Step1 Download the mobile app from this site Download Android App from here Step 2: Open the app and search using the zip location or the place of interest. (Due to disk space constraint we have truncated analytical data in MySQL to have limited values)

Step 3: User is presented with Bar and pie charts that show the crime rates and crime types for the searched location. The app also shows the rank of the place in terms of safety.

(pie chart showing the crime analytics)

(bar chart showing the crime analytics) Based on the Graphs plotted with 10 years of data and breakup if the crime rates and crime types user can make a conscious decision if the locality is safe or not. Based on the Big data analytics in previous steps , the mobile client application also tells user if the crime rate has increased or decreased on the searched location.

Roadmap for future

This POC is only a feasibility check and proof of concept of the analytical abilities of Big Data platform and mobile application as the easiest way for various users to potentially make use of the analytical capabilities. From here we are planning to use this platform to have various other analytical models created which are as follows:

1. 1. 1. 1. Analytical model of Crime Rate of the searched location and its nearby places.

         2. Analytical model to use the crime rate to suggest most safe path in google maps from one location to another

         3. Use the income data and employment along with the crime data to identify pattern of categories of job most people in the neighborhood has.