The Positive Impact of Data Mining
The Positive Impact of Data Mining   

In the 5th Annual Survey (2011) data miners shared examples of situations where
data mining is having a positive impact on society.  The five areas mentioned most
often were:
  • Health / Medical Progress
  • Business Improvements
  • Personalized Communications & Marketing
  • Fraud Detection
  • Environmental

Below are selected "positive impact examples" the data miners shared.  Complete
lists are also available by following the links below.


Health / Medical Progress:

Thirty-six data miners participating in the 5th Annual Survey described positive
impact examples in this area.  Medical topics mentioned by several respondents
included:  drug and treatment discovery, improved diagnosis,
treatment
compliance, and healthcare optimization / cost reduction.   (All 36 health / medical
responses
can be seen here.)

Selected examples

  • Fast and accurate diagnosis of infectious diseases in patients admitted
    to ER unit.  Reference:  "Classification of infectious diseases based on
    chemiluminescent signatures of phagocytes in whole blood" by Daria
    Prilutsky, M.Sc.; Boris  Rogachev, M.D.; Robert S Marks, Ph.D.; Leslie  
    Lobel, M.D., Ph.D.; Mark Last, Ph.D., to appear in Artificial Intelligence
    in Medicine.

  • Created an improved health risk model tuned to our specific population
    that helps employees identify their personal risk of chronic disease
    (CVD, diabetes, cancer), then recommends actions that they can take
    to avoid them.  See http://cathealthbenefits.cat.com/cda/layout?
    m=164481&x=7 for an example.

  • More than 90% of the chemicals on the market have not been tested on
    their potential hazards on humans and environment like
    carcinogenicity, developmental toxicity, mutagenicity, skin
    sensitization.  If they would be tested in traditional way, i.e., by animal
    testing, it would require more than 10 million additional animals, billions
    of euro, and more than 50 years in time.  Mathematical models (qsar
    models) obtained from observation data are going to help reducing,
    refining, and replacing animal tests in the near future.

  • Using analytics to identify health plan members that are non-compliant
    in keeping their diseases under control (e.g. diabetes). That way the
    health plan can send educational literature/programs to those most
    likely to be out of compliance and keep them healthier..


Business Improvements:

Eighteen data miners participating in the 5th Annual Survey described positive
impact examples in this area.  (
All 18 business improvement responses
can be seen here.)

Selected examples:

  • We have been using data mining techniques to improve electricity
    network forecasts which in turn help manage network capital     
    expenditure costs which in turn reduces electricity bills.  We have also
    been developing new tariff scenarios and identifying potentially energy
    vulnerable groups within the community.

  • Using DM to develop a model to optimize the scheduling of oil shipping
    tankers (reduced cost to the company and positive environmental
    impact).  Due to client confidentiality agreements I couldn't share the
    details of the project.

  • Determining patterns associated with successful call center agents and
    applying those learnings in recruitment and training of agents led to
    increased employee, organization and customer satisfaction.  That was
    fulfilling.

  • We have used data mining for employee attrition. This will generate
    long term ROI of millions of dollars.

  • Predictive Maintenance: the ability to maintain a machine or prevent
    accidents from happening. Predictive Maintenance could have prevent
    the tragedy in the Gulf of Mexico.


Personalized Communications & Marketing:

Eighteen data miners participating in the 5th Annual Survey described positive
impact examples in this area.  (
All 18 responses can be seen here.)

Selected examples:

  • As we are working mainly for a marketing audience, we have proven to
    make marketing more efficient (+10% revenue) and more relevant
    (+300% more relevant) by applying predictive analytics for targeting
    and customized offering.

  • Netflix recommender and Google Reader's "sort by magic" have
    completely revolutionized the way I consume information for
    entertainment and news.

  • Automation of keyword pricing for bidding on search engines.  
    Automation of "good" keyword detection.

  • Amazon customer experience embodies data mining.

  • In a world where people are inundated with marketing, data mining can
    create an environment that a consumer hears what they want to hear
    about.

  • Discover information in text.  Predict who is likely to buy, migrate or
    churn.  Improve campaigns.

  • Data mining can help quality of life and assist public health
    interventions - by producing a model  predicting the effect of an
    intervention.  Data mining can help filter important data and information
    from spam and information overload.


Fraud Detection:

Thirteen data miners participating in the 5th Annual Survey described positive
impact examples in this area.  (All 13 responses can be seen here.)

Selected examples:

  • Insurance Fraud Bureau uses link analytics and data mining within
    Detica NetReveal to detect motor fraud .  Up to date over 150 arrests
    have been made on organised criminal rings and seized millions of
    pounds in proceeds of crime reducing premium costs for us all.

  • Insurance fraud detection and insurance risk estimation - better prices
    for better clients.

  • Preventing financial transaction fraud.


Environmental:

Nine data miners participating in the 5th Annual Survey described positive impact
examples in this area.  (
All 9 responses can be seen here.)

Selected examples:

  • We are development model for:  -Predict the flow of rivers to meet two
    months before the availability of irrigation water and the potential for
    flooding and / or drought.  -Predict the occurrence of hail  from a  
    weather radar data.  -Predict crop yield and production from monthly
    variables surveyed in the field.  -Sort prevailing production systems to
    perform an economic analysis of production system and not for each
    particular activity.

  • Reduction of postharvest losses through better ability to call which lines
    of product to commit to which storage / distribution channels.

  • More than 90% of the chemicals on the market have not been tested on
    their potential hazards on humans and environment like
    carcinogenicity, developmental toxicity, mutagenicity, skin
    sensitization.  If they would be tested in traditional way, i.e., by animal
    testing, it would require more than 10 million additional animals, billions
    of euro, and more than 50 years in time.  Mathematical models (qsar
    models) obtained from observation data are going to help reducing,
    refining, and replacing animal tests in the near future.


Other Positive Impact Areas:

Data miners participating in the 5th Annual Survey described positive impact
examples in a very diverse set of areas, including crime detection, rhino horn
fingerprinting, student retention, charity donor appeals, substance abuse treatment,
genetics, pollution reduction, politics / government, risk management, cost
optimization, poverty /economic planning, combustion optimization, petroleum
discovery
, and more.  (All 44 "other" responses can be seen here.)

Selected examples:

  • Better prediction than classical statistics methods for rhino horn
    fingerprinting.   Can help busy people develop improved prediction
    models that can handle non-linearity and complex interactions.

  • We have applied data mining in government to better estimate the
    social and economical return of funding programs (loans, grants, and
    so on) using credit scoring and data mining in general.  By our
    estimations, these techniques translate into more jobs created, more
    funding programs, and in general a virtuous circle.

  • I work for a charity, which is using data mining to increase the level of
    donation to appeals, reduce attrition in regular giving and identify likely
    candidates for major giving, bequests etc.

  • Data mining offers tools that can improve our knowledge about risk
    factors in adolescent substance use.

  • In our case, predicting academic performance help us detect at risk
    students in order to help them, so they will not quit their studies.  We   
    will also detect the high performance students to get the best of them.
Copyright (c) 2012 Rexer Analytics, All rights reserved