cisc 849 : applications in fintech namami shukla dept of computer & information sciences...

22
CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based Banking Customer Analytics

Upload: ralph-walker

Post on 17-Jan-2018

215 views

Category:

Documents


0 download

DESCRIPTION

CISC 849 : Applications in Fintech Every digital process and social media exchange produces it. Systems, sensors and mobile devices transmit it.

TRANSCRIPT

Page 1: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Namami ShuklaDept of Computer & Information Sciences

University of Delaware

iCARE : A Framework for Big Data Based Banking Customer Analytics

Page 2: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Big Data

• Extremely Large Datasets that may be analysed computationally to reveal patterns ,trends and associations.

• Three major characteristics of Big Data.

Page 3: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

• Every digital process and social media exchange produces it.

• Systems, sensors and mobile devices transmit it.

Page 4: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Customer Analytics

•Give organizations the customer insight necessary to deliver

offers that are anticipated, relevant and timely.

•Decrease attrition by accurately predicting customers most likely

to leave.

•Developing the right proactive campaigns to retain them.

Page 5: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Challenges

• To handle the massive amount of complex data in a cost-

effective and efficient way.

• To effectively generate business value from the analytics and

obtain competitive advantages for banks.

Page 6: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Intelligent Customer Analytics for the Recognition and Exploration (iCARE)

framework.

•It is presented to analyze banking customer behaviors from banking big data.

•IBM products - IBM SPSS* Analytic Server and IBM InfoSphere BigInsights* are used.

•Analytical models are Customized and validated on the processed data according to specified business scenarios

iCARE Framework

Page 7: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

• BigInsights platform is an Apache Hadoop based hardware-agnostic software platform.

• IBM SPSS Analytic Server (AS) provides big data analytics capabilities.

• It also includes integrated support for unstructured predictive analytics from the Hadoop environment.

Page 8: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Architecture

Page 9: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

There are four phases in the solution:

•Data acquisition •Data preparation•Data modeling•Business applications

iCare Solution Design

Page 10: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

• A standard input format is defined in iCARE for structured data.

• Unstructured data can originate from inside a bank, including web log files, call records and external resources.

• The unstructured data is usually stored as files rather than database tables.

Data Acquisition

Page 11: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

• To enhance the data quality both unstructured and structured data needs additional data preparation.

• Big Sql tool is used to efficiently handle the incomplete, incorrect, or irrelevant data.

o Use of statistical methods

• Data integration

Data Preparation

Page 12: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

• Integrated data is stored in a data warehouse.

• Data Conflict is resolved. • Consolidated enterprise customer view is generated.

Page 13: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

The iCARE analytical models can be built for different business scenarios based on business objectives.

Advantages :

•All the statistic and machine learning methods are customized to suit corresponding business scenarios.

• Use of Parallelized models

iCARE Analytical Models

Page 14: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Example Traditional K-means clustering algorithm.

•To divide n data points into K clusters with similar points to minimize the total distance between the points to their cluster centers.

Steps :

•The algorithm first randomly selects K data points to be the cluster centers.

•Then assigns each data point to the closest cluster, and updates the center of each cluster by calculating weighted average of all data points in the cluster.

Page 15: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

A customized and parallelized K-means clustering algorithm.

It is used to segment customers of a bank into several clusters based on their profile and transaction information. K data points are selected to be the cluster centers.

Steps:

• First data point is used as first cluster center.

•For each data point, compute the minimum distance between it and each defined cluster center.

Page 16: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

The Manhattan distance is used here. The Manhattan distance is used here.

• Use of manhattan distance .

• For two D-dimensional data points x and y, the metric is defined as

• New cluster center is selected.

• Each data point is assigned to the closest cluster using standard algorithm with distance metric.

• Updated cluster center :

Page 17: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

•Data points are redistributed to their closest cluster

•Parallelized further to follow map reduce model.

•The distances between data points and cluster centers are updated by the Mapper.

•Reducer adds up the partial sum to get new cluster center.

Page 18: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Business Applications

• Customer segmentation and preference analysis

• Potential customer identification

• Customer network analysis

• Market potential analysis

• Channel allocation and operation optimization

Page 19: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Conclusion

•It can be extended for other data analytics applications, not limited to customer relationship management or the banking industry.

•iCARE framework is scalable by adding more parallel analytical models.

Page 20: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

20 terabytes of data was analyzed to help generate insights for

retaining active online banking customers.

Implemented for commercial bank in southeast China

Start-up

Page 21: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

The model ran 12 times faster as a single host for the 4 GB test

data sample with 1,600 instances.

Page 22: CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based

CISC 849 : Applications in Fintech

Thank You !