Part 7: Model Appraisal.


117 views
Uploaded on:
Category: Food / Beverages
Description
7. Part 7: Model Evaluation. 7. Section 7: Model Evaluation. Appraisal Sorts. The Model Correlation instrument gives. C. KS. Outline insights Factual representation. ASE. 7. Section 7: Model Appraisal. Synopsis Measurements Outline. Expectation Sort. Measurement.
Transcripts
Slide 1

7 Chapter 7: Model Assessment

Slide 2

7 Chapter 7: Model Assessment

Slide 3

Assessment Types The Model Comparison instrument gives C KS Summary insights Statistical design ASE

Slide 4

7 Chapter 7: Model Assessment

Slide 5

Summary Statistics Summary Prediction Type Statistic Accuracy/Misclassification Profit/Loss KS-measurement Decisions ROC Index (concordance) Gini coefficient 1,2,3,… Rankings Average squared slip SBC/Likelihood p ≈ E( Y ) ^ Estimates

Slide 6

Summary Statistics Summary Prediction Type Statistic Accuracy/Misclassification Profit/Loss KS-measurement Decisions ROC Index (concordance) Gini coefficient 1,2,3,… Rankings Average squared mistake SBC/Likelihood p ≈ E( Y ) ^ Estimates

Slide 7

Summary Statistics Summary Prediction Type Statistic Accuracy/Misclassification Profit/Loss KS-measurement Decisions ROC Index (concordance) Gini coefficient 1,2,3,… Rankings Average squared lapse SBC/Likelihood p ≈ E( Y ) ^ Estimates

Slide 8

Comparing Models with Summary Statistics This show delineates the Model\'s utilization Comparison device, which gathers evaluation data from appended displaying hubs and empowers you to effortlessly think about model execution measures.

Slide 9

7 Chapter 7: Model Assessment

Slide 10

Decisions Sensitivity outlines Response rate diagrams Statistical Graphics Summary Prediction Type Statistic 1,2,3,… Rankings p ≈ E( Y ) ^ Estimates ...

Slide 11

Apply model to acceptance information. Factual Graphics – Prediction Ranks approval information ...

Slide 12

main 40% Prediction Ranks Select top n % cases. ...

Slide 13

Sensitivity-Based Plots Count portion of essential result cases in determination. main 40% 1.0 affectability 0.0 ...

Slide 14

False Positive Fraction Count division of optional result cases in determination. main 40% 1.0 affectability 0.0 1.0 false positive portion (1-specificity) ...

Slide 15

ROC Chart Repeat for all choice divisions. 1.0 affectability 0.0 1.0 false positive portion (1-specificity) ...

Slide 16

1.0 affectability 0.0 1.0 false positive division (1-specificity) ROC Index ROC Index (c-measurement) ...

Slide 17

1.0 0.5 0% 100% 40% percent chose (decile) Response Rate Charts main 40% Select top n % cases. ...

Slide 18

Cumulative Gain main 40% Count division of cases in choice with essential result. 1.0 aggregate increase 0.5 0% 100% 40% percent chose (decile) ...

Slide 19

Cumulative Gains Chart Repeat for all determination parts. 1.0 combined increase 0.5 0% 100% percent chose (decile) ...

Slide 20

Comparing Models with Statistical Graphics This show represents the utilization of measurable illustrations to think about models.

Slide 21

Adjusting for Separate Sampling This exhibition outlines how to conform for partitioned examining in SAS Enterprise Miner.

Slide 22

7 Chapter 7: Model Assessment

Slide 23

Outcome Overrepresentation The specimen size is resolved not by the aggregate number of cases but rather by the quantity of cases in minimum basic result (generally essential). ...

Slide 24

Separate Sampling Cases are examined independently from every result. Case: • test every single essential cas • coordinate every essential case by one or more auxiliary cases ...

Slide 25

Separate Sampling Benefit • Similar prescient force with littler case check

Slide 26

Separate Sampling Consequences • Must conform evaluation measurements and design • Must change expectation gauges for inclination

Slide 27

Adjusting for Separate Sampling (proceeded with) This exhibit delineates how to alter for discrete inspecting in SAS Enterprise Miner.

Slide 28

Creating a Profit Matrix This show outlines how to make a benefit lattice.

Slide 29

7 Chapter 7: Model Assessment

Slide 30

Profit Matrices request disregard 14.86 0 essential result - 0.68 0 auxiliary result 0 benefit conveyance for request choice

Slide 31

Profit Matrices request overlook 14.86 0 essential result - 0.68 0 optional result 0 benefit dispersion for request choice

Slide 32

pick the bigger ^ Expected Profit Solicit = 14.86 p 1 – 0.68 p 0 Expected Profit Ignore = 0 Decision Expected Profits request disregard 14.86 0 essential result - 0.68 0 optional result 0 ...

Slide 33

Decision Threshold request overlook 14.86 0 essential result - 0.68 0 auxiliary result 0 choice limit ^ p 1 ≥ 0.68/15.54  Solicit ^ p 1 < 0.68/15.54  Ignore

Slide 34

Average Profit request disregard 14.86 0 essential result - 0.68 0 optional result 0 normal benefit Average benefit = (14.86ï‚\' N PS – 0.68 ï‚\' N SS )/N PS = # requested essential result cases N SS = # requested optional result cases N = aggregate number of appraisal cases

Slide 35

Evaluating Model Profit This exhibition outlines seeing the outcomes of joining a benefit grid.

Slide 36

Viewing Additional Assessments This show represents a few different evaluations of conceivable hobby.

Slide 37

Optimizing with Profit (Optional) This show represents improving your model entirely on benefit.

Slide 38

Exercise 1 This activity strengthens the ideas examined already.

Slide 39

7 Chapter 7: Model Assessment

Slide 40

Assessment Tools Review Compare model outline insights and measurable representation. Model Comparison Create choice information; include earlier probabilities and benefit frameworks. Information Source Tune models with normal squared slip or suitable benefit framework. Displaying Tools

Slide 41

Assessment Tools Review Obtain implies and different insights on information source variables. StatExplore .:tslidese

Recommended
View more...