derak presentation

26
New Data Project Plan Anderson, Brykman, Gero, Lakhani, Matthews PREDICT 480 – SECTION 55

Upload: eric-gero

Post on 10-Feb-2017

99 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Social Media Project Plan

New Data Project PlanAnderson, Brykman, Gero, Lakhani, MatthewsPREDICT 480 SECTION 55

IntroductionCompany BackgroundDERAK is an analytical support team for a young credit card companyCompany is 5 years oldLatest venture is to consolidate data sources across the organization as well as collect data from mobile applications to gain customer loyaltyWhy did we select this company?The combination of credit card data, social media, GPS, mobile data, etc. offer exiting opportunities for modeling to give value to the customer

IntroductionSWOT AnalysisStrengthsCompany is becoming analytically mature (Stage III on Maturity Model)Diversity of products to spread riskWeaknessesNot yet a significant brand in Credit Card marketAs a new company, we have limited data from 2008 Financial CrisisOpportunitiesGain new insights on customers through mobile appUtilize data for predictive modelingExplore uses of GPS dataSell customer data to third party companiesThreatsCredit Card Market is mature with several key players. Gaining significant market share is difficult and expensiveCustomer privacy concerns and government regulations

Introduction3 Key Issues to Address with AnalyticsLimited understanding about our customerWe need to know more about our customer in order to offer them more value than the competitionCredit card default riskIncrease volume but maintain qualityFraudWe need to increase trust with the customer by protecting them financially

IntroductionDatabase and Data preparation planOur database and data preparation plan will help us solve this issues by:Helping us to gather more intimate data on our customer to better understand their shopping preferences, personal interests, and measure customer loyaltyWe plan to utilize the new data to identify possible natural clusters of our customers that may help use to identify segments that could carry higher probability of default riskThrough mobile applications, we can use GPS data to identify transactions that may be fraudulent in order to better protect our customers

Literature and Data Sources Overview

Document your information and data sources. Describe methods relevant to:Acquiring Storing MaintainingAccessing the data that you needCite references that you have used to guide your thinking about data sources and methods, and include these in the reference list a the end of the paper.

Acquiring DataCompany Databases1Demographic informationBanking plansCredit/Debit balances Transactional dataSocial media3TwitterHashtags@BrandingFacebookStatus updatesLikesContact Network

1 https://customers.microsoft.com/Pages/Download.aspx?id=13928/2 http://www.sitetechsystems.com/top-10-ways-to-use-gis-in-retail-banking/2 www.mmaglobal.com/files/mbankingoverview.pdf3 http://www.bearingpoint.com/ecomaXL/files/0615_WP_EN_Social_CRM_final_web.pdf

Mobile Phone Data2GPS LocationMobile client applicationMobile webShort Message Service (SMS)Contact Lists

7

Bank Mobile Social mediaData Integration

Central locationDealing with data issues and data preparationIntegration of results from various data sources into a central location for analysis1Integration of results from feeds (structured and unstructured)1

1 https://www.in.capgemini.com/resource-file-access/resource/pdf/A_Case_for_Enterprise_Data_Management_for_Banking.pdf

StorageConsiderations close to real time or longer term access1Governance for retention periods, data ownership and entry of new data sources2Maintain backups2 Accessible2Document sources and validations taking place2

1 http://www.oracle.com/us/products/middleware/data-integration/oracle-goldengate-realtime-access-2031152.pdf2 http://www.osfi-bsif.gc.ca/eng/docs/data_maint_ja06.pdf

Maintaining dataBacking up data1Deleting data based on established retention1Performance optimization21 http://www.osfi-bsif.gc.ca/eng/docs/data_maint_ja06.pdf2 http://www.oracle.com/technetwork/database/bi-datawarehousing/twp-bp-for-stats-gather-12c-1967354.pdf

AccessingAccess via graphical user interfaces (GUIs) through applications forStructured data or unstructured dataReal time or ex-post dataAccess forSystem administration and maintenanceValidation, editing and estimating (VEE)Analyzing and modelling processed data (Ultimate Goal)Audit access controls1

1 http://www.osfi-bsif.gc.ca/eng/docs/data_maint_ja06.pdf

Criteria Describe the systems and methods used to: Acquire dataStore dataMaintain data Access dataDescribe the infrastructure.Describe how these systems and methods work together.

Acquire dataPython and R languages will be used for data acquisition.Lightweight and flexible.Mature - industry standards.Multiple libraries available for data handling and analytics.Data acquisitionFacebook1 and Twitter2 provide APIs for accessPython supportedReturns JSON formatCompany dataYears of company credit card history available.

1 https://developers.facebook.com/docs/graph-api/using-graph-api/v2.52 https://dev.twitter.com/overview/api

Storing, Maintaining and AccessingOracle DBMS for internal dataPostgreSQL for analyticsSupports structured and unstructured data.Better performance than MongoDBPurchase history tied to customers maintained for 90 days.Data used for analytic models maintained for 90 days.Data anonymized and maintained for 2 years.Used for general trending and analysis.Sold to third parties.Views will be created in the PostgreSQL environment to allow access for the mobile application.

Infrastructure and How it works.

Data PreparationText AnalysisCleanse the data for keywords

GPS DataBroad scale demographic segmentation

Contact NetworksIdentifiable features of people and groups

Data IssuesProblemsApplication Permissions

Consumer Participation

Inaccurate Text Analysis

Data Storage and Processing

SolutionsProvide benefits for permissions

Entice consumers with offers

Segmented Analysis

Commodity Clustered Servers

Data QualityOutliers & Incomplete dataUsed where applicable, likely excluded

Bootstrapping Estimation of additional sampling distributions

Influential observationsUsed to further define possible segmentations

In ConclusionOpportunitiesDERAK is well positioned to create key analytical insights to our credit card customer, which will benefit its services by:Improving its market accessGaining new insights on its credit card customers through mobile appUtilizing the collected data for predictive modeling (spending habits, etc.)Exploring uses of GPS dataSelling anonymized customer data to third party companies

In ConclusionFull Services

In ConclusionData Sources

In ConclusionSystems and Methods

In ConclusionData Issues and Preparation

In ConclusionDirection and future stateRefining data collection and analyticsImproving data mining methodsExpanding our services and market reach to other industries

Team and Presenters (in order of appearance)Aaron MatthewsKetan LakhaniEric GeroDaniel AndersonRaphael Brykman

ReferencesRoyal Bank of Scotland Case Study (2013)Mobile Banking Overview (January 2009)10 ways to use GIS data in Retail Banking (2012)A Case for Enterprise Data Management in Banking (2012)Data Maintenance at IRB Institutions (2006)Best Practices for Gathering Optimizer Statistics with Oracle Database 12c