Machine learning for marketing of branded consumer products
A method including retrieving a product information from a database is provided. The method includes associating the product information with multiple classification values, forming a vector associated with a consumer product. The classification values form coordinates of the vector in a vector space that comprises multiple vectors associated with multiple consumer products. The method includes determining a cluster in the vector space, including at least one vector selected according to a relative distance within a cluster boundary. The method includes selecting a discriminator vector from a vector difference between a first vector in a first cluster in the vector space and a second vector in a second cluster in the vector space and identifying a new consumer product associated with a new vector that is formed by adding the discriminator vector to a third vector from the vector space, the third vector associated with a known consumer product.
1. A computer-implemented method comprising:
training, to simulate a shopping basket based on a plurality of classification values from a vector space, a first neural network, wherein the vector space comprises multiple vectors, each vector associated with a consumer product having at least one of the plurality of classification values;
wherein training the first neural network comprises selecting a discriminator vector from a vector difference between a first vector in a first cluster in the vector space and a second vector in a second cluster in the vector space, wherein the discriminator vector is indicative of a direction in the vector space that classifies a desirable attribute of the consumer product;
training the first neural network according to a consumer data from a selected consumer information or from an aggregated information from a consumer population;
training, to detect whether or not a shopping basket is simulated or exist in a retailer database, a second neural network;
determining, with the second neural network, whether the shopping basket from the first neural network passes a threshold value, the threshold value indicative of whether a shopping basket is simulated or not; modifying the first neural network when the shopping basket passes the threshold value; and modifying the second neural network when the shopping basket fails the threshold value.
2. The computer-implemented method of claim 1 , wherein modifying one of the first neural network comprises re-enforcing a model coefficient in the first neural network, and modifying the second neural network comprises re-enforcing a model coefficient in the second neural network.
3. The computer-implemented method of claim 1 , further comprising determining a responsiveness of an aggregated group of consumers to the consumer product by verifying an inclusion of the consumer product in the shopping basket for each consumer in the aggregated group of consumers.
4. The computer-implemented method of claim 1 , further comprising determining a change in a responsiveness of an aggregated group of consumers to a first consumer product by verifying an inclusion of a second consumer product in the shopping basket for each consumer in the aggregated group of consumers.
5. The computer-implemented method of claim 1 , wherein training the first neural network or the second neural network comprises determining a change of desirability for a first product in the shopping basket by adjusting a product feature of a second product in the shopping basket.
6. The computer-implemented method of claim 1 , wherein training the first neural network according to a consumer data comprises identifying a modified consumer product associated with a new vector that is formed by adding or subtracting a discriminator vector to a third vector from the vector space, the third vector associated with a known consumer product, wherein the modified consumer product is selected from a test basket.
7. The computer-implemented method of claim 1 , wherein modifying the first neural network comprises providing a signal to the first neural network and determining a change in the shopping basket responsive to the signal, wherein the signal includes at least one of a coupon a bargain or a promotion associated with one or more consumer products.
8. The computer-implemented method of claim 1 , wherein modifying the first neural network comprises providing a signal to the first neural network and determining a change in the shopping basket responsive to the signal, wherein the signal includes one of an environmental factor, a geographic factor, or a demographic factor.
9. The computer-implemented method of claim 1 , wherein modifying the first neural network comprises providing a signal to the first neural network and determining a change in the shopping basket responsive to the signal, wherein the signal includes a substitution of a first product with a second product in the shopping basket.
10. A system, comprising:
one or more processors; and
a memory storing instructions which, when executed by the one or more processors, cause the system to perform operations, comprising to:
train, to simulate a shopping basket based on a plurality of classification values from a vector space, a first neural network, wherein the vector space comprises multiple vectors, each vector associated with a consumer product having at least one of the plurality of classification values;
wherein to train the first neural network the one or more processors execute instructions to select a discriminator vector from a vector difference between a first vector in a first cluster in the vector space and a second vector in a second cluster in the vector space, wherein the discriminator vector is indicative of a direction in the vector space that classifies a desirable attribute of the consumer product;
train the first neural network according to a consumer data from a selected consumer information or from an aggregated information from a consumer population;
train, to detect whether or not a shopping basket is simulated or exist in a retailer database, a second neural network;
determine, with the second neural network, whether the shopping basket from the first neural network passes a threshold value, the threshold value indicative of whether a shopping basket is simulated or not;
modify the first neural network when the shopping basket passes the threshold value; and modify the second neural network when the shopping basket fails the threshold value.
11. The system of claim 10 , wherein modifying one of the first neural network comprises re-enforcing a model coefficient in the first neural network, and modifying the second neural network comprises re-enforcing a model coefficient in the second neural network.
12. The system of claim 10 , wherein the one or more processors further execute instructions to determine a responsiveness of an aggregated group of consumers to the consumer product and verify an inclusion of the consumer product in the shopping basket for each consumer in the aggregated group of consumers.
13. The system of claim 10 , wherein the one or more processors further execute instructions to determine a change in a responsiveness of an aggregated group of consumers to a first consumer product and verify an inclusion of a second consumer product in the shopping basket for each consumer in the aggregated group of consumers.
14. The system of claim 10 , wherein to train the first neural network or the second neural network the one or more processors execute instructions to determine a change of desirability for a first product in the shopping basket by adjusting a product feature of a second product in the shopping basket.
15. The system of claim 10 , wherein to train the first neural network according to a consumer data the one or more processors execute instructions to identify a modified consumer product associated with a new vector that is formed by adding or subtracting a discriminator vector to a third vector from the vector space, the third vector associated with a known consumer product, wherein the modified consumer product is selected from a test basket.
16. The system of claim 10 , wherein to modify the first neural network the one or more processors execute instructions to provide a signal to the first neural network and to determine a change in the shopping basket responsive to the signal, wherein the signal includes at least one of a coupon a bargain or a promotion associated with one or more consumer products.
17. The system of claim 10 , wherein to modify the first neural network the one or more processors execute instructions to provide a signal to the first neural network and determining a change in the shopping basket responsive to the signal, wherein the signal includes one of an environmental factor, a geographic factor, or a demographic factor.
18. The system of claim 10 , wherein to modify the first neural network the one or more processors execute instructions to provide a signal to the first neural network and determining a change in the shopping basket responsive to the signal, wherein the signal includes a substitution of a first product with a second product in the shopping basket.