Activity for Svetlana Levitan
-
Svetlana Levitan committed [a2aa8b]
Add stringLength to Built-in functions and accuracy, AUC, F1, etc. to ModelExplanation, Mantis issues 181 and 256
-
Great! I am glad you found a solution. What scoring engine are you using? Have a nice day! Svetlana From: "Dennis" dehe92@users.sourceforge.net To: "[pmml:discussion] " 187860@discussion.pmml.p.re.sourceforge.net Date: 04/01/2020 02:54 PM Subject: [EXTERNAL] [pmml:discussion] Additional output with information about invalid/missing Values (Multiple Models) Thanks a lot (also for the additional tips)! I had to do some changes (e.g. <compoundpredicate booleanoperator="or"> didn't work for me; I now...
-
Hi Dennis, Thank you for the good question. How do you plan to score your PMML? It is possible that different scoring engines may have different levels of support for some of the PMML constructs you are using. In particular, functionName="mixed" and Segmentation multipleModelMethod="selectAll" can be not very widely supported. In addition, I think your TreeModel has a problem: please add a second child to the root node, it can have <true> for predicate, but without it the tree does not look like...
-
Fix errors in the example of anomaly detection model based on isolation forest
-
Fix errors in the scoring of anomaly detection model based on isolation forest
-
Fix some issues in Anomaly Detection model
-
Fix a typo in absolute value in SVM example
-
Fix error in logistic Regression example
-
Update occurences of 4.3 to 4.4 in GeneralStructure document
-
Issue 236: all numberOf should be xs:nonNegativeInteger
-
Issues 168, 216
-
Issue 229: missing type of some attributes
-
Issue 226: allow aggregates on lag operator
-
Issue 235: use xs:double instead of xs:float in AssociationRule and KohonenMap attributes
-
Issue 230 - clarify parts of Time Series
-
Issue 215: allow additional models for predictors, fix a sign in a formula
-
Issue 218 Error in BN calculation
-
Issue 221: specify optype where possible for built in functions
-
Issue 8: Adding ARIMA and other models to Time Series
-
Issue 217: clarify handling of missing in built-in functions
-
Restore change for issue 212 lost when checked in change for 203. GaussianProcess.
-
Issue 170: Word separator by default is any number of spaces rather than one.
-
Issue 214: extraneous piece in Scorecard example
-
Issue 207: remove mention of division by 0, as this is covered in the document earlier
-
Hi Vinay, You are right, this looks like a typo. Let me open a Mantis issue about it, we should fix it in the next release. Very sorry for the big delay, apparently this page is not monitored well.
-
Hi Leslie, In my previous replies I explained how to return multiple values. You have to use OutputField element for each of those fields you want to get, with each having feature="transformedValue" and a MapValues elements having the mappings from the one tree predicted value to your output. Please see http://dmg.org/pmml/v4-3/Output.html and http://dmg.org/pmml/v4-3/Transformations.html#xsdElement_MapValues for details of those. Regarding the concatenation question, PMML does not allow you to do...
-
Issue 209: add attribute inferenceType to Bayesian Network model
-
Issue 213: default type of numeric Constant should be double
-
Issue 211: clarify calculation of missing models
-
Issue 203: added standardDeviation into the list of possible outputs, use it in GaussianProcess example
-
Issue 207: add function modulo
-
Issue 200: add displayName to ParameterField
-
Issue185: isValid, isNotValid
-
Issue 212: Gaussian Process variance formula
-
Issues 158 and 194, Bayesian networks
-
Issue 192: adding returnInvalid option for missingValueTreatment
-
Issue 206: adding isScorable to ExampleModel, fixing a typo in text
-
Issues 64, 160, 178, 197, 201
-
Issue 167: Add Extension into BlockIndicator
-
Issue 187: clarify definitions of outliers and invalid values in MiningSchema
-
Issue 196: clarify that ordinal targets are treated like categorical in KNN
-
Issue 193: Scorecard: change type of name attribute in Characteristic
-
Issue 199: add Value into OutputField to allow order specification for ordinal fields
-
Issue 166: clarify targets in Cox regression
-
Issues 175 and 202, MiningSchema
-
Restore added Extension in ClusteringModelQuality lost when change for 186 was checked in
-
Issue 182: wrong count of mining functions
-
Issue 195: allow Item elements to refer to Derived fields
-
Changes for issue 186, made by Andy
-
Issue 163: new built-in functions
-
Issue 177: Extension must be the first element
-
Issues 161 and 164
-
Add definition of algorithmName, issue 155
-
Minor corrections in Gaussian Process document
-
Fix some symbols in Regression
-
Issues 139 and 35
-
Issue 75
-
Issue 67
-
Issue 75
-
Issue 72
-
Add BayesinNetwork to General Structure
-
Issue 131
-
Issues 147, 149, 151, 152
-
Issue 146 Scoring for binomial distribution in ...
-
Issue #145: New activation function for neural...
-
Issue #87 / #86 / #91 Mining Model
-
Issue #148: OutputField needs an attribute ind...
-
Issue 136: Review method=indicator attribute in...
-
Issue 142: Field scope
-
Make small examples quoted
-
Fix some GP problems
-
Fix errors in BuiltInFunctions
-
Issue 144: Add built-in functions for cdf and p...
-
Issue137
-
Issue 78: Add maxOccurs=unbounded to LiftData
-
Issue 94: allow PROB-NUMBER to accept scientifi...
-
Issue 143: make dataType of OutputField required
-
Issue 138: add CategoricalPredictor and maxWins...
-
Issues 134 and 135: TextIndex element
-
latest from 4.2 branch
-
Issue 129: remove sentence about EXPRESSION
-
Issue 141: Gaussian process
-
Hi Jose: The only way to implement what you want that I can see is this: you use...
-
Hi Jose: I am sorry that my explanation was not very clear. Yes, I meant that you...
-
Hi Jose: In your example there is a one-to-one correspondence between results for...
-
Hi Jose: In addition to the predicted value (stored as "score" attribute of each...
-
Sorry, that email was not finished and not intended to be sent out. My apologies....
-
Hi Priit: Thank you for the interesting question. Another disadvantage of LOESS is...
-
clarify when lowValue and highValue are needed
-
Use latest PREHTML from 4.2 branch
-
Hi Debashis: Thank you for the good question. I think the reason for this avoidance...
1
