change point detection in insurance portfolios
play

Change-point detection in insurance portfolios Laub Work in - PowerPoint PPT Presentation

Change-point detection in insurance portfolios Dr Patrick J. Change-point detection in insurance portfolios Laub Work in progress Introduction Problem Data Dr Patrick J. Laub ML Options Post-doctoral researcher with Chaire DAMI BNP


  1. Change-point detection in insurance portfolios Dr Patrick J. Change-point detection in insurance portfolios Laub Work in progress Introduction Problem Data Dr Patrick J. Laub ML Options Post-doctoral researcher with Chaire DAMI BNP Paribas Cardif Technical Seminar March 27, 2019 1 / 17

  2. Who am I Change-point detection in insurance Mathematician + portfolios Software Engineer ≈ Dr Patrick J. Laub Data Scientist Introduction Problem PhD in Applied Data Probability between ML Options University of Queensland (Brisbane) and Aarhus University Arrived to Lyon in September PhD Thesis 2 / 17

  3. Since arriving Re / submitted 3 papers Change-point detection in insurance portfolios R e s i D d e n c e e P Database access p e r m a i t r t m Hosting Visiting Professors e n Dr Patrick J. t S Apartment hunting é m i n Laub a i r e Bank account Introduction CPAM Préfecture Server / GPU access Problem s Data Teaching w Reporting e ML Options i v r e French classes t Surgery n Séminaire Lyon - Lausanne I Conference submissions 3 / 17

  4. Since arriving: Paper accepted Change-point detection in insurance portfolios Dr Patrick J. Laub Introduction Problem Data ML Options Shortened recording of a seminar on the topic: https://youtu.be/8Ih2NxrLrmg 3 / 17

  5. Since arriving: Created and delivered a course on rare-event estimation/simulation Change-point detection in insurance portfolios Dr Patrick J. Laub Introduction Problem Data ML Options Lecture notes & recordings at: https://pat-laub.github.io/rare-events/ 3 / 17

  6. The BNP Netherlands contracts tables Change-point detection in insurance portfolios Dr Patrick J. Laub C 2014 = Introduction Problem Data ML Options C 2015 = 4 / 17

  7. The problem Change-point detection in insurance portfolios Did our risk exposure change in some significant way? ⇒ We can update premiums/reserves. Dr Patrick J. Laub Introduction Look at Problem C t C t +1 and Data given C 0 , C 1 , . . . , C t +1 and throw up an alarm if ML Options distance ( C t , C t +1 ) > threshold . Restriction: Don’t assume some parametric form of C t . 5 / 17

  8. Prior work Change-point Nothing has been done on exactly this topic... detection in insurance portfolios Similar to “fraud detection” in the sense that the events of Dr Patrick J. Laub interest are rare... but otherwise no similarity. Introduction Problem Standard change-point detection is a huge field, mostly Data parametric (e.g. hidden Markov models, Markov regime ML Options switching models). Mostly uni- or low-dimensional. Require samples to be piecewise stationary/i.i.d. and to keep a fixed dimension. Similar to the discriminator in a GAN; that tries to learn a distribution and discern perturbations from the learned distribution. 6 / 17

  9. Data: Big Picture Change-point Number of payouts made across the different coverage types detection in insurance portfolios Dr Patrick J. Laub Introduction Problem Data ML Options 7 / 17

  10. Data: Big Picture Change-point Total payouts made across the different coverage types detection in insurance portfolios Dr Patrick J. Laub Introduction Problem Data ML Options 7 / 17

  11. Data: Big Picture Change-point detection in insurance portfolios Claim outcomes Dr Patrick J. Laub Introduction Problem Data ML Options 7 / 17

  12. Data: Big Picture Change-point detection in insurance portfolios Claim number distribution Dr Patrick J. Laub Introduction Problem Data ML Options 7 / 17

  13. Data: Big Picture Change-point detection in insurance portfolios Claim number distribution (zoomed) Dr Patrick J. Laub Introduction Problem Data ML Options 7 / 17

  14. All Columns Change-point Contracts table Claims table detection in insurance policy-nr, coverage-type, insured-id, product-code, portfolios broker-nr, agency-nr, network-nr, certificate-nr, professional-status, contract-status, coverage-status, Dr Patrick J. last-coverage-status-change, premium-type, loss-number, application-code, gender, birth-date, Laub premium-schema, commission-schema, status-eng, policy-number, prod-code, product, commission-type, birth-date, gender, effect-date, dm, insured-risk, insured-amount, start-date-contract, Introduction currency, percentage-premium-increase, policy-duration, end-date-contract, end-date-benefit, claim-waiting-period, type-of-loan, date-of-loss-occurence, date-of-loss-declaration, Problem percentage-annuity-decreasing, date-loss-was-closed, type-of-loss, franchise-abs, insured-amount–capital, insured-amount-type, decision, toekenningsreden, Data fiscality, premium-increase, incasso-machtiging, p, reason-commercial-coulance, coulance-monitoring, pm, percentage-chosen-commission, beroep, total-amount-paid-euro, first-payment, ML Options percentage-chosen-commission-y1, benefit-period-of-first-payment, latest-payment, percentage-chosen-commission-y2, benefit-period-of-last-payment, percentage-chosen-commission-y3, smoking, number-of-paid-periods, relatienummer, death-coverage-type, ao-coverage-type, verz-kapitaal, invoerdat, oorzaak-type, ao-vaststelling, claim-payment-duration, reden-van-afwijzing, type-lening, ao-percentage, child-addition, repatriation, dm1, premium-dm1, ao-vaststelling-, juridisch-proces-date, reopened-date, premium-dm2, payment-received, uitkeringspercentage, pemsa-insuredid, last-paid-billing-date, single-com, rev-com, pemsa-coverage-type, coverage-duration, surrenderdate, invoerdatum, discount-amount, pemsa-product-code discount-dm, discount-type, coverageendreason, tariff, pawntaker, occupation, occupation-code, occupation-class 8 / 17

  15. All Columns − Unused Change-point Contracts table Claims table detection in insurance policy-nr, coverage-type, insured-id, product-code, portfolios broker-nr, agency-nr, network-nr, / certificate-nr, / / / / / / / / / / / / / / professional-status, / contract-status, coverage-status, / / / / / / / / / / / / / / / / Dr Patrick J. last-coverage-status-change, premium-type, loss-number, application-code, gender, birth-date, Laub / premium-schema, commission-schema, / / / / / / / / / / / / / / / / / / / status-eng, policy-number, prod-code, product, / commission-type, birth-date, gender, effect-date, dm, / / / / / / / / / / / / / / / / / / / insured-risk, insured-amount, start-date-contract, Introduction currency, percentage-premium-increase, policy-duration, end-date-contract, end-date-benefit, claim-waiting-period, type-of-loan, date-of-loss-occurence, date-of-loss-declaration, Problem / percentage-annuity-decreasing, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / date-loss-was-closed, type-of-loss, franchise-abs, insured-amount–capital, / insured-amount-type, / / / / / / / / / / / / / / / / / / / / / / / / decision, / / / / / / / / / toekenningsreden, / / / / / / / / / / / / / / / / / / / / Data fiscality, premium-increase, incasso-machtiging, p, reason-commercial-coulance, coulance-monitoring, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / pm,/ percentage-chosen-commission, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / beroep, total-amount-paid-euro, first-payment, ML Options / percentage-chosen-commission-y1, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / benefit-period-of-first-payment, latest-payment, / percentage-chosen-commission-y2, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / benefit-period-of-last-payment, / percentage-chosen-commission-y3, smoking, / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / number-of-paid-periods,/ relatienummer, / / / / / / / / / / / / / / / / / death-coverage-type, ao-coverage-type, verz-kapitaal, invoerdat, oorzaak-type, ao-vaststelling, / claim-payment-duration, / / / / / / / / / / / / / / / / / / / / / / / / / / / / reden-van-afwijzing, type-lening, ao-percentage, / / / / / / / / / / / / / / / / / / / / / / child-addition, repatriation, dm1, premium-dm1, ao-vaststelling-, / juridisch-proces-date, reopened-date, / / / / / / / / / / / / / / / / / / / / / / / premium-dm2, payment-received, uitkeringspercentage, pemsa-insuredid, last-paid-billing-date, single-com, rev-com, pemsa-coverage-type, coverage-duration, surrenderdate, invoerdatum,/ discount-amount, / / / / / / / / / / / / / / / / / / / pemsa-product-code / discount-dm,/ / / / / / / / / / / / / / / discount-type, / / / / / / / / / / / / / / / / coverageendreason, / / / / / / / / / / / / / / / / / / / / / tariff,/ pawntaker, / / / / / / / / / / / / / occupation, / / / / / / / / / / / / / occupation-code, / / / / / / / / / / / / / / / / / / / occupation-class / / / / / / / / / / / / / / / / / / 8 / 17

Recommend


More recommend