Systems and methods for mining model accuracy display for multiple state prediction
AssigneeMicrosoft Corporation
InventorsZhaohui Tang, Pyungchul Kim
Life of this patent
Estimating the term…
Abstract
Systems and methods are provided for producing a mining model accuracy display that depicts the model's accuracy at predicting a state for a multiple-state variable. The model predicts a state and provides an associated probability for each case. Points are graphed such that one coordinate of the data point corresponds to a number N of cases and the other coordinate corresponds to the number of correct predictions made in the top N cases by probability.
Claims28
A method for accuracy display for a data mining model, where said model predicts, for each case in a testing data set, a specific state from among a state set of at least two possible states and where said model determines an associated probability for each of said predictions, and where the correctness of said prediction is verifiable, comprising:
- producing a visual representation of a multi-state prediction evaluation on a graph of at least two dimensions comprising at least two data points where for each of said data points:
- (1) a data point set of said cases from among said testing data set has been selected such that no case in said data point set has a lower associated probability for the prediction made by said model than any case not in said data point set, (2) a first coordinate of said data points corresponds to the number of cases in said data point set, and (3) a second coordinate of said data points corresponds to the number of correct predictions for cases contained in said data point set.
A computer-readable medium comprising computer-executable modules having computer-executable instructions for accuracy display for a data mining model, where said model predicts, for each case in a testing data set, a specific state from among a state set of at least two possible states and where said model determines an associated probability for each of said predictions, and where the correctness of said prediction is verifiable, said modules comprising:
- a module for producing a visual representation of a multi-state prediction evaluation on a graph of at least two dimensions comprising at least two data points where for each of said data points:
- (1) a data point set of said cases from among said testing data set has been selected such that no case in said data point set has a lower associated probability for the prediction made by said model than any case not in said data point set, (2) a first coordinate of said data points corresponds to the number of cases in said data point set, and (3) a second coordinate of said data points corresponds to the number of correct predictions for cases contained in said data point set.
A computer device for accuracy display for a data mining model, where said model predicts, for each case in a testing data set, a specific state from among a state set of at least two possible states and where said model determines an associated probability for each of said predictions, and where the correctness of said prediction is verifiable, comprising:
- means for producing a visual representation of a multi-state prediction evaluation on a graph of at least two dimensions comprising at least two data points where for each of said data points:
- (1) a data point set of said cases from among said testing data set has been selected such that no case in said data point set has a lower associated probability for the prediction made by said model than any case not in said data point set, (2) a first coordinate of said data points corresponds to the number of cases in said data point set, and (3) a second coordinate of said data points corresponds to the number of correct predictions for cases contained in said data point set.
Dependencies are inferred from the claim text.