Systems and methods for determining substitutions

A system comprising:

• one or more processors;
• and
• one or more non-transitory computer-readable storage devices storing computing instructions configured to run on the one or more processors and cause the processors to perform:
• determining when an item of a list of items is out of stock;
• and
• when the item is out of stock:
• determining a list of possible substitutes for the item of the list of items, the list of possible substitutes comprising a set of items in an item catalogue;
• determining respective qualities for each possible substitute of the list of possible substitutes;
• training a machine learning algorithm on labeled training data comprising (1) all positive historical acceptance data and (2) randomly sampled negative historical acceptance data, wherein each labeled training datum of the labeled training data is labeled with at least one of:
• a respective probability of a respective similarity between a respective historical title for a respective historical item and a respective historical description of a respective historical substitute;
• a respective taxonomy difference for the respective historical item and the respective historical substitute;
• a respective price difference between the respective historical item and the respective historical substitute;
• or
• a respective normalized sales rank comparison for the respective historical item and the respective historical substitute;
• determining a respective similarity score for each possible substitute of the list of possible substitutes using the machine learning algorithm, as trained;
• determining a respective historical substitution score for each possible substitute of the list of possible substitutes;
• determining a respective final score for each possible substitute of the list of possible substitutes using the respective similarity score for each possible substitute of the list of possible substitutes, the respective historical substitution score for each possible substitute of the list of possible substitutes, and at least one or more rectifiers having ReLU non-linearity that enables training of deep supervised neural networks without unsupervised pre-training;
• ranking each possible substitute of the list of possible substitutes based upon the respective final score for each possible substitute;
• facilitating a display, on a user interface of a user device, of a highest ranked possible substitute of the list of possible substitutes;
• receiving, from the user interface of the user device, a selection of the highest ranked possible substitute of the list of possible substitutes;
• after receiving the selection of the highest ranked possible substitute, substituting the highest ranked possible substitute of the list of possible substitutes for the item of the list of items;
• storing the selection of the highest ranked possible substitute as additional training data with the labeled training data;
• and
• re-training the machine learning algorithm on the additional training data and the labeled training data.