Model Data¶
(Activity) for Tier: Data Analytics
PURPOSE¶
This activity creates specific abstract models of available observational data that have predictive and/or explanatory value in answering the question under study.
WHEN¶
Data Exploration exit criteria met.
PARTICIPATING ROLES¶
- ACCOUNTABLE
- RESPONSIBLE
ENTRY CRITERIA¶
- Prepared data met data exploration exit criteria.
SUB-ACTIVITIES¶
- Select a specific modeling technique to apply. If multiple techniques are potentially applicable, perform this activity separately for each one. When selecting a technique identify any assumptions the technique makes about data and ensure those assumptions apply in the available data.
- Generate test design. Document the intended plan to train, test, and evaluate the model. This must include how the data will be divided into training, test, and validations sets when applicable to the selected technique. Include decisions made on the number of iterations or other steps required.
- Build model. Determine and set the initial parameters for the model and document the rationale for those parameters. Generate model using the selected technique and perform any required post-processing. Document a description of the model, revised parameter settings, guidance on interpretation, and any other considerations about how model should be used.
- Assess model. Evaluate model results using evaluation criteria. Rank results of alternative models as applicable. Interpret results in context of and using terms from problem domain. Consider plausibility and reliability of results. Decide on success or failure of technique.
OUTPUTS¶
- Source code and data structures applicable to the selected technique along with the description of the model and its use.
Todo
need to define work products for all outputs.
EXIT CRITERIA¶
- One or more models execute successfully against validation data or a determination is made none is possible.