Combined wide and deep machine learning models for automated database element processing systems, methods and apparatuses
A method of automated database element processing includes training a wide machine learning model with historical feature vector inputs to generate a wide ranked element output. The method includes training a deep machine learning model with the historical feature vector inputs to generate a deep ranked element output. The method includes generating a set of inputs specific to an individual entity, obtaining a set of current article database elements, and creating a feature vector input according to the set of inputs and the set of current article database elements. The method includes processing the feature vector input with the wide machine learning model to generate a wide ranked element list, processing the feature vector input with the deep machine learning model to generate a deep ranked element list, and merging database elements of the wide and deep ranked element lists to generate a ranked element recommendation output.
1. A computer-based recommendation system comprising:
at least one computer readable non-transitory memory storing computer-executable instructions including at least one trained wide machine learning model configured to generate a wide ranked article output and at least one trained deep machine learning model configured to generate a deep ranked article output; and
at least one processor coupled with the at least one computer readable non-transitory memory and that performs the following operations upon execution of the computer-executable instructions:
generating a set of inputs specific to a subscriber and derived from article access records associated with the subscriber;
obtaining a set of current articles published according to specified time period criteria;
creating a feature vector input according to the set of inputs specific to the subscriber and the set of current articles;
generating, via the at least one trained wide machine learning model, a wide ranked article list from a wide model subset of the current articles generated based on the feature vector input and based on a subscriber click history, and having a highest correlation with the article access records;
generating, via the at least one trained deep machine learning model, a deep ranked article list from a deep model subset of the current articles generated based on the feature vector input and based on the subscriber click history, and having a highest correlation with the article access records;
generating a ranked article recommendation output by merging articles from the wide ranked article list and the deep ranked article list; and
displaying the ranked article recommendation output on a user device of the subscriber.
2. The computer-based recommendation system of claim 1 , wherein the set of current articles comprises news articles.
3. The computer-based recommendation system of claim 1 , wherein the set of current articles comprises newspaper articles.
4. The computer-based recommendation system of claim 1 , wherein at least one of the trained wide machine learning model and the trained deep machine learning model is trained on the subscriber click history associated with historical articles.
5. The computer-based recommendation system of claim 4 , wherein the subscriber click history comprises a number of clicks.
6. The computer-based recommendation system of claim 4 , wherein the subscriber click history comprises a click history since a last training of the at least one of the trained wide machine learning model and the trained deep machine learning model.
7. The computer-based recommendation system of claim 4 , wherein both the trained wide machine learning model and the trained deep machine learning model are trained on the subscriber click history associated with historical articles.
8. The computer-based recommendation system of claim 1 , wherein the operations further include obtaining the subscriber click history over a time period.
9. The computer-based recommendation system of claim 8 , wherein the time period is at least 45 days.
10. The computer-based recommendation system of claim 1 , wherein at least one of the trained wide machine learning model and the trained deep machine learning model trained on the subscriber click history associated with historical articles is a supervised trained model.
11. The computer-based recommendation system of claim 1 , wherein at least one of the trained wide machine learning model and the trained deep machine learning model trained on the subscriber click history associated with historical articles is an unsupervised trained model.
12. The computer-based recommendation system of claim 1 , wherein the specified time period criteria comprises an elapsed time requirement since the trained wide machine learning model and the trained deep machine learning model were trained.
13. The computer-based recommendation system of claim 12 , wherein the specified time period criteria comprises a sliding window.
14. The computer-based recommendation system of claim 12 , wherein the elapsed time requirement comprises a time period threshold.
15. The computer-based recommendation system of claim 14 , wherein the time period threshold is at least 45 days.
16. The computer-based recommendation system of claim 15 , wherein the time period threshold is at least six months.
17. The computer-based recommendation system of claim 1 , wherein the operations further include retraining or finetuning the trained wide machine learning model and the trained deep machine learning model.
18. The computer-based recommendation system of claim 17 , wherein the trained wide machine learning model and the trained deep machine learning model are retrained or finetuned periodically.
19. The computer-based recommendation system of claim 18 , wherein the trained wide machine learning model and the trained deep machine learning model are retrained or finetuned every hour.
20. The computer-based recommendation system of claim 18 , wherein the trained wide machine learning model and the trained deep machine learning model are retrained or finetuned every five minutes.
21. A computer-based recommendation method comprising:
generating a set of inputs specific to a subscriber and derived from article access records associated with the subscriber;
obtaining a set of current articles published according to specified time period criteria;
creating a feature vector input according to the set of inputs specific to the subscriber and the set of current articles;
generating, via at least one trained wide machine learning model configured to generate a wide ranked article output, a wide ranked article list from a wide model subset of the current articles generated based on the feature vector input and based on a subscriber click history, and having a highest correlation with the article access records;
generating, via at least one trained deep machine learning model configured to generate a deep ranked article output, a deep ranked article list from a deep model subset of the current articles generated based on the feature vector input and based on the subscriber click history, and having a highest correlation with the article access records;
generating a ranked article recommendation output by merging articles from the wide ranked article list and the deep ranked article list; and
displaying the ranked article recommendation output on a user device of the subscriber.
22. A non-transitory computer-readable medium storing computer-executable instructions, the computer-executable instructions comprising:
generating a set of inputs specific to a subscriber and derived from article access records associated with the subscriber;
obtaining a set of current articles published according to specified time period criteria;
creating a feature vector input according to the set of inputs specific to the subscriber and the set of current articles;
generating, via at least one trained wide machine learning model configured to generate a wide ranked article output, a wide ranked article list from a wide model subset of the current articles generated based on the feature vector input and based on a subscriber click history, and having a highest correlation with the article access records;
generating, via at least one trained deep machine learning model configured to generate a deep ranked article output, a deep ranked article list from a deep model subset of the current articles generated based on the feature vector input and based on the subscriber click history, and having a highest correlation with the article access records;
generating a ranked article recommendation output by merging articles from the wide ranked article list and the deep ranked article list; and
displaying the ranked article recommendation output on a user device of the subscriber.