Predictive Analytics Modeling
10 likes1,310 views
The document discusses predictive analytics and its application in decision-making, highlighting the importance of data modeling, simulation, and optimization in retaining employees and improving recruitment. It covers various statistical methods such as A/B testing, decision trees, and regression, emphasizing the need for data-driven leadership. The presentation concludes with a Q&A section and provides contact information for further engagement.
Predictive Analytics Modeling
- 1. Predictive Analytics: Modeling Concepts April 15th, 2014 Presented by: Andrew Pulvermacher Director | Predictive Analytics in/drewpulvermacher Â
- 2. 2  âWeâre  forge+ng  how  to  ďŹy.â  -ÂâRory  Kay,  United  captain  and  co-Ââchairman  of  a  Federal  AviaAon  AdministraAon  commiCee  âamong  the  accidents  and  certain  categories  of  incidents  that  were  examined,  roughly  two-Ââthirds  of  the  pilots  either  had  diďŹculty  manually  ďŹying  planes  or  made  mistakes  using  ďŹight  computers.â     This  reliance  on  computer-Ââheavy  ďŹight  decks  and  the  âproblems  that  result  when  crews  fail  to  properly  keep  up  with  changes  in  levels  of  automaAonâ  now  pose  âthe  biggest  threatsâ  to  airliner  safety  in  the  world. Â
- 3. 3 Â
- 5. 5    Forward-ÂâLooking  Decision  Making    ObjecAve  |  Variables  |  Constraints  TODAY:  Modeling  Concepts  Predictive Analytics Series 1.⯠ExecuAve  IntroducAon  2.⯠Data  Modeling  3.⯠SimulaAon  4.⯠OpAmizaAon  5.⯠Data-ÂâDriven  Leadership Â
- 7. in/drewpulvermacher  Predictive Analytics Modeling Concepts WHAT is flying the planeâŚ
- 8. 8  OBJECTIVE:  We  need  to  RETAIN  our  top  employees  and  RECRUIT  More  EďŹecAvely  Starts  with  Data  Easy  3  Step  Process  1)⯠Data  2)⯠What  Column  3)⯠Select  The  TECHNOLOGY  does  it  for  me  Recommenda@on:  Give  everyone  a  PosiAve  Performance  Review  (?) Â
- 9. in/drewpulvermacher  Predictive Analytics Modeling Concepts 1.âŻClassiďŹcaAon  2.âŻA/B  TesAng  3.âŻDecision  Trees  4.âŻRegression Â
- 10. 10  10  ClassiďŹca@on  1)⯠Hidden  RelaAonships  2)⯠Data  Familiarity  3)⯠SegmentaAon Â
- 11. 11  11  A/B  Tes@ng  In  markeAng,  A/B  tes@ng  is  a  simple  randomized  experiment  with  two  variants,  A  and  B,  which  are  the  control  and  treatment  in  the  controlled  experiment.    It  is  a  form  of  sta@s@cal  hypothesis  tes@ng.  Â
- 12. 12  12  A/B  Tes@ng  Customer  Response  Rate  Failure  Rate  Pricing  Loan  Default Â
- 13. 13  Summary Statistics Probability  of  Breast  Cancer:  0.8%  â˘âŻ Woman  >  40  MAMMOGRAM  Unknown:  Breast  Cancer  YES  NO  Test  Posi@ve  90%  7%  Flesh  &  Blood  Example Â
- 14. 14 Â BAYES Â THEOREM Â 1 Â 70 Â
- 15. 15  Decision  Trees  Reason  for  Being:  Contingent Probabilities Noun:  the  probability  that  an  event  will  occur  given  that  one  or  more  other  events  have  occurred Â
- 16. 16  Decision  Trees:  Scenario  PRODUCT  A  Prob($)  =  60%  PRODUCT  B  Prob($)  =  20%  Prob($)  =  40% Â
- 17. 17  Letâs  Plant  a  (Decision)  Tree  3  Probabili@es  of  Success  for  Product  B  Results  are  NOT  Independent.  p(B)  =  p(B|A)  x  p  (A)  +  p(B  |  notA)  x  p(not  A) Â
- 18. 18  One  More?  Regression Â
- 19. 19  âAs  our  machines  get  faster  and  ingest  more  data,  we  allow  ourselves  to  be  dumber.   Instead  of  wrestling  with  our  problems  in  earnest,  we  can  just  plug  in  billions  of  examples  of  them.    Which  is  a  bit  like  using  a  graphing  calculator  to  do  your  high-Ââ school  calculus  homework  â  it  works  great  un@l  you  need  to  actually  understand  calculus.â     -Ââ  James  Somers,  âThe  Man  Who  Would  Teach  Machines  to  Think,â      The  AtlanAc,  November  2013 Â
- 21. in/drewpulvermacher  Predictive Analytics Modeling Concepts 1.âŻClassiďŹcaAon  2.âŻA/B  TesAng  3.âŻDecision  Trees  4.âŻRegression Â
- 23. Thank you for attending our webinar 23  Drew@PerformanceG2.com " âŻCall us: 877.742.4276 " ⯠ Email us: info@performanceg2.com or drew@performanceg2.com " ⯠ Visit our web site: performanceg2.com " ⯠ Read our Analytics blog: performanceg2.com/blog " ⯠ Follow us: " âŻ(Twitter) @performanceg2 " âŻ(Facebook) /performanceg2 " âŻ(YouTube) /performanceg2 " âŻ(LinkedIn) /performanceg2-inc
- 24. Predictive Analytics Series 1.⯠ExecuAve  IntroducAon  2.⯠Data  Modeling  3.⯠SimulaAon  â  May  15,  2014  4.⯠OpAmizaAon  5.⯠Data-ÂâDriven  Leadership  Visit  hCp://www.performanceg2.com/ events  to  register  for  the  upcoming  series. Â