TIMi modeler training sessions

There are 2 different courses:

For simple users

This course is open to any profile: Business, IT, dataminer.
Duration: 2 to 3 days.

TIMi Modeler is really easy to use for anybody. For a day-to-day usage, the usage of TIMi Modeler is quite straight forward. The videos available on this page covers most of the standard “basic” usage of TIMi.

However, a short course covering the different aspects of the modelization process might still be good idea to take:

  • What kind of problem can you solve with TIMi Modeler? Cross-selling models, probability-of-default models, churn models, propensity models, fraud models, risk models, segmentation models.
    The predictive models for “Strategic & Tactic decisions” are more complex and will be explained inside the “expert user course”.
  • What methodological steps do I have to follow when working with TIMi Modeler? Do I still need to check if the distributions of my variables are normal before starting the modelization?
  • How to avoid the traps linked to the time periods? How to define properly the target (i.e. the target group), taking into account the time dimension?
  • This topic will be discussed very briefly and on request only: What are the algorithms in TIMi Modeler? What are the limitations of these algorithms? How to avoid these limitations? How does these algorithms relates to other algorithms available in other datamining/ML packages? What are the theoretical mathematical properties of these algorithms that ensure that TIMi Modeler consistently delivers better models than all the other statistical softwares (like SAS, KXEN, ART)?
  • How to get the most out of the reports generated with TIMi Modeler? How to give a business sense to the charts and tables inside the TIMi Modeler reports? (with a special session on models for optimal allocation of resources.)
  • What are the most important parameters of TIMi Modeler? How do these parameters relate to the general theory of modeling? How to fine tune these parameters?
  • What to do when you have no or nearly no lift? How to boost the quality of your model?
  • How to handle different kind of Modelization problems? What are the best variables to create or use inside your dataset in:
    • a Telco application?
    • a predictive maintenance application?
    • a Cross-Selling Banking application?
    • a Credit-Risk Banking application?
    • a classical CRM application?
  • What are the traps and how to avoid them when…
    • …creating continuous models?
    • …combining multiple binary models into one N-ary model?
  • What strange behavior could appears out of the modelization process and how to avoid them?

For expert users

This course is open to any profile that is interested in the more mathematical aspect behind TIMi Modeler : Business, IT, dataminer.
Duration: 1 to 2 days + simple user course.

This course covers the following subjects:

  • Predictive models for “Strategic & Tactic decisions”.
  • What are the different “advanced parameters” of TIMi Modeler ?
    How to use them to increase the accuracy of the predictive models created with TIMi Modeler?
  • What are the different techniques available to construct models? What are the weakness of each modelization techniques? How to overcome these weakness?
  • What’s overfitting? How to prevent it?
    • Pruning
    • Ridge regression
    • Backward Stepwise
    • Forward Stepwise
    • Cross-validation
  • What are the danger linked to sampling?
  • Combination of models: Boosting, Bagging, Feature Selection