Data analysis apparatus, data analysis method, and data analysis program

A data analysis apparatus comprising:

• a processor configured to execute a machine learning program including a learning process and a prediction process, wherein, in the learning process, learning parameters are calculated in such a manner that an error between a predicted value obtained from a feature vector variable that is an input and a true value is minimum and, in the prediction process, a new predicted value is calculated from data not used for learning which are test data;
• and
• a storage device that stores the program, wherein
• the processor is configured to execute, within the learning process:
• a selection process for selecting, from a set of feature variables that are training data, a first feature variable group that is a trivial feature variable group, wherein trivial feature variables are nearly equivalent to the true value of the predicted value, and a second feature variable group with feature variables other than the first feature variable qroup;
• an operation process for varying a first regularization coefficient related to a first weight parameter group corresponding to the first feature variable group among a set of weight parameters configuring a prediction model in such a manner that a loss function is larger, and varying a second regularization coefficient related to a second weight parameter group corresponding to the second feature variable group among the set of weight parameters configuring the prediction model in such a manner that the loss function is smaller, wherein the loss function relates to a difference between a prediction result output in a case of inputting the set of feature variables to the prediction model and ground truth data corresponding to the feature variables;
• and
• a weight parameter learning process for learning the set of weight parameters of the prediction model in such a manner that the loss function is minimum as a result of varying a first regularization coefficient and a second regularization coefficient by the operation process;
• and
• the processor is further configured to execute, within the prediction process, calculating the prediction result and prediction accuracy of the prediction result based on the prediction result and the ground truth data, and
• in the operation process, the processor is configured to re-operate the first regularization coefficient and the second regularization coefficient in a case in which the prediction accuracy is equal to or lower than predetermined prediction accuracy, and
• in the weight parameter learning process, the processor is configured to re-learn the set of weight parameters of the prediction model in such a manner that the loss function is minimum as a result of re-operating the first regularization coefficient and the second regularization coefficient by the operation process.