System and method for generating recommendations with cold starts
A system and method generate recommendations with cold starts. The system comprises a hardware-based processor, a memory, and a set of modules. The memory stores item taxonomy data for at least one item, stores client descriptive data for at least one client, and stores historical response data for at least one item responded to by the at least one client. The set of modules includes a machine learning module and a recommendation module. The machine learning module generates a response probability matrix using the historical response data, the item taxonomy data, and the client descriptive data. The recommendation module generates and outputs a recommendation corresponding to input data using the information deduced in the response probability matrix. The method implements the system.
1 . A recommendation system, comprising:
an input device configured to receive input data including an existing client feature corresponding to an existing client, and a new client feature corresponding to a new client;
a hardware-based processor;
a memory configured to store instructions, configured to provide the instructions to the hardware-based processor, and configured to store item taxonomy data for at least one item, to store client descriptive data for the existing client, and to store historical response data for at least one item responded to by the existing client, wherein the item taxonomy data includes an item feature of an existing item, wherein the client descriptive data includes the existing client feature of the existing client, and wherein the memory does not store historical response data for the new client; and
a set of modules configured to implement the instructions provided to the hardware-based processor, the set of modules including:
a machine learning module configured to generate a response probability matrix using the historical response data, the item taxonomy data, and the client descriptive data, wherein the response probability matrix includes a probability of a response of the existing client or the new client to the at least one item, and wherein the probability of the response in the response probability matrix in the case of the new client is generated using the machine learning module performing a first similarity comparison of the existing client feature of the existing client in the client descriptive data to the new client feature of the new client; and
a recommendation module configured to generate and output a relevant recommendation of the at least one item corresponding to input data of the existing client or the new client using the response probability matrix, wherein in the case of the new client, the recommendation module generates and outputs the relevant recommendation of the at least one item for the new client without any historical response data for the new client, thereby performing a cold start for generating and outputting the relevant recommendation for the new client.
2 . The recommendation system of claim 1 , wherein the input data includes item data.
3 . The recommendation system of claim 2 , wherein the item data is selected from the group consisting of: a numerical value, an idea, a process, a product, a service, an application, and a financial value.
4 . The recommendation system of claim 1 , wherein the machine learning module includes a neural network.
5 . The recommendation system of claim 4 , wherein the neural network is a deep neural network including a plurality of nodes arranged in a plurality of layers of nodes.
6 . The recommendation system of claim 4 , wherein the neural network is trained from a training set including the historical response data, the item taxonomy data, the client descriptive data, and probabilities as target outputs associated with the historical response data, the item taxonomy data, and the client descriptive data.
7 . The recommendation system of claim 1 , wherein the machine learning module performs the first similarity comparison using a first threshold.
8 . The recommendation system of claim 1 , wherein the memory does not initially store historical response data for the new client.
9 . The recommendation system of claim 1 , wherein the at least one item is a new item.
10 . A recommendation system, comprising:
an input device configured to receive input data including an existing client feature corresponding to an existing client, and a new client feature corresponding to a new client;
a hardware-based processor;
a memory configured to store instructions, configured to provide the instructions to the hardware-based processor, and configured to store item taxonomy data for at least one item, to store client descriptive data for the existing client, and to store historical response data for at least one item responded to by the existing client, wherein the item taxonomy data includes an item feature of an existing item, wherein the client descriptive data includes the existing client feature of the existing client, and wherein the memory does not store historical response data for the new client; and
a set of modules configured to implement the instructions provided to the hardware-based processor, the set of modules including:
a trained machine learning module configured to generate a response probability matrix using the historical response data, the item taxonomy data, and the client descriptive data, wherein the response probability matrix includes a probability of a response of the existing client or the new client to the at least one item, and wherein the probability of the response in the response probability matrix in the case of the new client is generated using the trained machine learning module performing a first similarity comparison of the existing client feature of the existing client in the client descriptive data to the new client feature of the new client; and
a recommendation module configured to generate and output a relevant recommendation of the at least one item corresponding to the input data of the existing client or the new client using the response probability matrix, wherein in the case of the new client, the recommendation module generates and outputs the relevant recommendation of the at least one item for the new client without any historical response data for the new client, thereby performing a cold start for generating and outputting the relevant recommendation for the new client.
11 . The recommendation system of claim 10 , wherein the input data includes item data.
12 . The recommendation system of claim 11 , wherein the item data is selected from the group consisting of: a numerical value, an idea, a process, a product, a service, an application, and a financial value.
13 . The recommendation system of claim 10 , wherein the machine learning module includes a neural network.
14 . The recommendation system of claim 13 , wherein the neural network is a deep neural network including a plurality of nodes arranged in a plurality of layers of nodes.
15 . The recommendation system of claim 13 , wherein the neural network is trained from a training set including the historical response data, the item taxonomy data, the client descriptive data, and probabilities as target outputs associated with the historical response data, the item taxonomy data, and the client descriptive data.
16 . The recommendation system of claim 10 , wherein the trained machine learning module performs the first similarity comparison using a first threshold.
17 . The recommendation system of claim 10 , wherein the memory does not initially store historical response data for the new client.
18 . The recommendation system of claim 10 , wherein the at least one item is a new item.
19 . A computer-based method, comprising:
providing a hardware-based processor, a memory, and a set of modules, wherein the memory is configured to store instructions, configured to provide the instructions to the hardware-based processor, and configured to store item taxonomy data for at least one item, to store client descriptive data for an existing client, and to store historical response data for at least one item responded to by the existing client, and the set of modules is configured to implement the instructions provided to the hardware-based processor, the set of modules including a trained machine learning module and a recommendation module;
receiving input data including an existing client feature corresponding to the existing client, and a new client feature corresponding to a new client;
storing, in the memory, item taxonomy data for at least one item wherein the item taxonomy data includes an item feature of an existing item, client descriptive data for the existing client wherein the client descriptive data includes the client feature of the existing client, and historical response data for the at least one item responded to by the existing client, and wherein the memory does not store historical response data for the new client;
generating a response probability matrix by the trained machine learning module using the historical response data, the item taxonomy data, and the client descriptive data, wherein the response probability matrix includes a probability of a response of the existing client or the new client to the at least one item, including:
in the case of the new client, performing a first similarity comparison, using the trained machine learning module, of the existing client feature of the existing client in the client descriptive data to the client feature of the new client; and
in the case of the new client, generating the probability of the response in the response probability matrix for the new client using the first similarity comparison;
receiving a recommendation request for a relevant recommendation of the at least one item for the existing client or the new client;
generating, using the recommendation module, the relevant recommendation corresponding to the recommendation request using the response probability matrix, wherein in the case of a new client, generating the relevant recommendation of the at least one item for the new client is performed without any historical response data for the new client, thereby performing a cold start for generating the relevant recommendation for the new client; and
outputting the relevant recommendation from the recommendation module.
20 . The computer-based method of claim 19 , wherein the machine learning module is trained from a training set including the historical response data, the item taxonomy data, the client descriptive data, and probabilities as target outputs associated with the historical response data, the item taxonomy data, and the client descriptive data.