Meta-feature training models for machine learning algorithms
A system and method are described herein for training and evaluating machine learning models trained using large complex datasets in an efficient manner. The systems enable methods for training a first model and generating predictions using the first trained model. The predictions and a portion of the data may be used to train a second model and generate predictions as an output of the second model. The first model may be used on a third dataset to generate predictions, which are appended to the third dataset. The appended third dataset is then input into the second model for generation of additional outputs that are compared against the third dataset for accuracy.
1. A method for developing and training a neural network, the method comprising:
receiving a dataset comprising text data and non-text data;
determining, based at least in part on the text data and the non-text data, that the dataset is to be split into a first number of datasets;
splitting the dataset into the first number of datasets that includes at least a first dataset, a second dataset, and a third dataset, the splitting comprising: populating the first dataset only with the text data, populating the second dataset only with the non-text data, and populating the third dataset with a combination of the text data and the non-text data;
training a first model using the first dataset based on a first set of features;
generating a first prediction for each of a plurality of first data using the first model and based on the first set of features;
augmenting the second dataset with the first prediction to obtain an augmented second dataset;
training a second model using the augmented second dataset based on a second set of features;
providing the augmented second dataset as input to the first model;
generating a second prediction for text data of a plurality of second data using the first model based at least in part on receiving the augmented second dataset as input and based at least in part on the second set of features;
augmenting the third dataset with the second prediction to obtain an augmented third dataset; and
generating a final prediction for each item in the augmented third dataset using the second model, wherein the augmented third dataset includes the text data of the plurality of second data, the non-text data of the plurality of second data and the second prediction for text data of the plurality of second data using the first model, wherein an accuracy of the final prediction is higher than an accuracy of each of the first prediction and the second prediction, and wherein the accuracy associated with using the first number of datasets is higher than a second accuracy associated with using a different plurality of datasets having a different number of datasets than the first number.
2. The method of claim 1 further comprising splitting a complete dataset into the received dataset, the plurality of first data, and the plurality of second data.
3. The method of claim 2 , wherein splitting the complete dataset comprises randomly dividing the complete dataset into equal groupings for the received dataset, the plurality of first data, and the plurality of second data.
4. The method of claim 1 , wherein the first model and the second model are different types of models.
5. The method of claim 1 , wherein the first model and the second model are a same type of model with different parameters.
6. The method of claim 1 , wherein the received dataset, the plurality of first data, and the plurality of second data are each the same size.
7. The method of claim 1 , wherein the received dataset is a different size than at least one of the plurality of first data or the plurality of second data.
8. A non-transitory computer-readable medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to:
receive a dataset comprising text data and non-text data;
determine, based at least in part on the text data and the non-text data, that the dataset is to be split into a first number of datasets;
split the dataset into the first number of datasets that includes at least a first dataset, a second dataset, and a third dataset, the splitting comprising: populating the first dataset only with the text data, populating the second dataset only with the non-text data, and populating the third dataset with a combination of the text data and the non-text data;
train a first model using the first dataset based on a first set of features;
generate a first prediction for each of a plurality of first data using the first model and based on the first set of features;
augment the second dataset with the first prediction to obtain an augmented second dataset;
train a second model using the augmented second dataset based on a second set of features;
provide the augmented second dataset as input to the first model;
generate a second prediction for text data of a plurality of second data using the first model based at least in part on receiving the augmented second dataset as input and based at least in part on the second set of features;
augment the third dataset with the second prediction to obtain an augmented third dataset; and
generate a final prediction for each item in the augmented third dataset using the second model, wherein the augmented third dataset includes the text data of the plurality of second data, the non-text data of the plurality of second data and the second prediction for text data of the plurality of second data using the first model, wherein an accuracy of the final prediction is higher than an accuracy of each of the first prediction and the second prediction, and wherein the accuracy associated with using the first number of datasets is higher than a second accuracy associated with using a different plurality of datasets having a different number of datasets than the first number.
9. The non-transitory computer-readable medium of claim 8 , further comprising additional instructions that, when executed by the one or more processors, cause the one or more processors to:
split a complete dataset into the received dataset, the plurality of first data, and the plurality of second data.
10. The non-transitory computer-readable medium of claim 9 , wherein splitting the complete dataset comprises randomly dividing the complete dataset into equal groupings for the received dataset, the plurality of first data, and the plurality of second data.
11. The non-transitory computer-readable medium of claim 8 , wherein the first model and the second model are different types of models.
12. The non-transitory computer-readable medium of claim 8 , wherein the first model and the second model are a same type of model with different parameters.
13. The non-transitory computer-readable medium of claim 8 , wherein the received dataset, the plurality of first data, and the plurality of second data are each the same size.
14. The non-transitory computer-readable medium of claim 8 , wherein the received dataset is a different size than at least one of the plurality of first data or the plurality of second data.
15. A system comprising:
a processor; and
a non-transitory computer-readable medium having instructions stored thereon that, when executed by the processor, cause the processor to:
receive a dataset comprising text data and non-text data;
determine, based at least in part on the text data and the non-text data, that the dataset is to be split into a first number of datasets;
split the dataset into the first number of datasets that includes at least a first dataset, a second dataset, and a third dataset, the splitting comprising: populating the first dataset only with the text data, populating the second dataset only with the non-text data, and populating the third dataset with a combination of the text data and the non-text data;
train a first model using the first dataset based on a first set of features;
generate a first prediction for each of a plurality of first data using the first model and based on the first set of features;
augment the second dataset with the first prediction to obtain an augmented second dataset;
train a second model using the augmented second dataset based on a second set of features;
provide the augmented second dataset as input to the first model;
generate a second prediction for text data of a plurality of second data using the first model based at least in part on receiving the augmented second dataset as input and based at least in part on the second set of features;
augment the third dataset with the second prediction to obtain an augmented third dataset; and
generate a final prediction for each item in the augmented third dataset using the second model, wherein the augmented third dataset includes the text data of the plurality of second data, the non-text data of the plurality of second data and the second prediction for text data of the plurality of second data using the first model, wherein an accuracy of the final prediction is higher than an accuracy of each of the first prediction and the second prediction, and wherein the accuracy associated with using the first number of datasets is higher than a second accuracy associated with using a different plurality of datasets having a different number of datasets than the first number.
16. The system of claim 15 , wherein the computer-readable medium comprises further instructions that, when executed by the one or more processors, cause the one or more processors to:
split a complete dataset into the received dataset, the plurality of first data, and the plurality of second data.
17. The system of claim 16 , wherein splitting the complete dataset comprises randomly dividing the complete dataset into equal groupings for the received dataset, the plurality of first data, and the plurality of second data.
18. The system of claim 15 , wherein the first model and the second model are different types of models.
19. The system of claim 15 , wherein the received dataset is a different size than at least one of the plurality of first data or the plurality of second data.
20. The system of claim 15 , wherein the first model and the second model are a same type of model with different parameters.