US 12,762,477 B2Granted

Machine learning model generation for time dependent data

AssigneeOracle International Corporation

InventorsVikas Agrawal, Karthik Bangalore Mani, Krishnan Ramanathan

Life of this patent

Estimating the term…

Loading patent…

Abstract

Embodiments generate a machine learning (“ML”) model. Embodiments receive training data, the training data including time dependent data and a plurality of dates corresponding to the time dependent data. Embodiments date split the training data by two or more of the plurality of dates to generate a plurality of date split training data. For each of the plurality of date split training data, embodiments split the date split training data into a training dataset and a corresponding testing dataset using one or more different ratios to generate a plurality of train/test splits. For each of the train/test splits, embodiments determine a difference of distribution between the training dataset and the corresponding testing dataset. Embodiments then select the train/test split with a smallest difference of distribution and train and test the ML model using the selected train/test split.

Claims20

Dependent claims

A method of generating a machine learning (ML) model comprising a target variable, the method comprising:

receiving training data, the training data comprising historical time dependent data, the historical time dependent data comprising a plurality of data points of transaction records, wherein each data point has a corresponding date when the data point occurred;
date splitting the training data by each of two or more of predetermined date fields of the plurality of corresponding dates to generate a plurality of candidate date split training data;
for each of the candidate date split training data, generating a plurality of train/test splits using a plurality of different train/test ratios;
for each of the plurality of train/test splits, generating a delay percentile vector for a corresponding training dataset and a corresponding testing dataset, the delay percentile vector comprising percentile values of target variable delay results for the corresponding dataset, and determining a difference of distribution between the corresponding training dataset and the corresponding testing dataset;
determining, from among the plurality of candidate date split training data and the plurality of train/test ratios, an optimized date split and an optimized train/test split having a smallest distribution of difference;
and
training the ML model using the optimized date split and optimized train/test split.

4 more elements

A non-transitory computer readable medium having instructions stored thereon that, when executed by one or more processors, cause the processors to generate a machine learning (ML) model comprising a target variable, the generating comprising:

receiving training data, the training data comprising historical time dependent data, the historical time dependent data comprising a plurality of data points of transaction records, wherein each data point has a corresponding date when the data point occurred;
date splitting the training data by each of two or more of predetermined date fields of the plurality of corresponding dates to generate a plurality of candidate date split training data;
for each of the candidate date split training data, generating a plurality of train/test splits using a plurality of different train/test ratios;
for each of the plurality of train/test splits, generating a delay percentile vector for a corresponding training dataset and a corresponding testing dataset, the delay percentile vector comprising percentile values of target variable delay results for the corresponding dataset, and determining a difference of distribution between the corresponding training dataset and the corresponding testing dataset;
determining, from among the plurality of candidate date split training data and the plurality of train/test ratios, an optimized date split and an optimized train/test split having a smallest distribution of difference;
and
training the ML model using the optimized date split and optimized train/test split.

4 more elements

A cloud based machine learning (ML) model generating system, the ML model comprising a target variable, the system comprising:

one or more processors executing instructions and configured to:
receive training data, the training data comprising historical time dependent data, the historical time dependent data comprising a plurality of data points of transaction records, wherein each data point has a corresponding date when the data point occurred;
date split the training data by each of two or more of predetermined date fields of the plurality of corresponding dates to generate a plurality of candidate date split training data;
for each of the candidate date split training data, generate a plurality of train/test splits using a plurality of different train/test ratios;
for each of the plurality of train/test splits, generate a delay percentile vector for a corresponding training dataset and a corresponding testing dataset, the delay percentile vector comprising percentile values of target variable delay results for the corresponding dataset, and determine a difference of distribution between the corresponding training dataset and the corresponding testing dataset;
determine, from among the plurality of candidate date split training data and the plurality of train/test ratios, an optimized date split and an optimized train/test split having a smallest distribution of difference;
and
training the ML model using the optimized date split and optimized train/test split.

5 more elements

Dependencies are inferred from the claim text.