System and method for predicting customer behavior
Various implementations of the invention for predicting customer behavior are described. Various implementations of the invention comprise an embedding component configured to receive and embed sequential inputs regarding a plurality of customer interactions with an online presence of a client; a plurality of causal dilated convolutional “CDC” elements configured to receive the embedded sequential inputs and to output a feature vector, where each CDC element comprises two causal dilated convolutions with regularization that is bypassed with a skip connection; a plurality of dense neural network elements configured to receive the feature vector and non-sequential inputs regarding a plurality of other customer interactions with the client, where each of the plurality of dense neural network elements comprises two dense neural networks with regularization that is bypassed with a skip connection; and an output generator configured to receive the output from the plurality of dense neural network elements and to generate a distribution of times over which a particular customer event will occur and/or a likelihood estimation that the particular customer event will occur within a particular time period.
1. A system for predicting customer behavior, the system comprising:
at least one computing processor;
an embedding component operably configured on the at least one computing processor to receive and embed sequential inputs regarding a plurality of customer interactions of a particular customer with an online presence of a client, wherein the plurality of customer interactions comprise event type data, product data, and source data, wherein each of the plurality of customer interactions further comprise a timing of that interaction;
a plurality of causal dilated convolutional “CDC” elements arranged in series, wherein the plurality of CDC elements is operably configured on the at least one computing processor to receive the embedded sequential inputs, wherein the plurality of CDC elements is configured to output a feature vector, wherein each of the plurality of CDC elements is configured to consecutively apply: a first causal dilated convolution, a first activation, a second causal dilated convolution, a second activation, and a skip connection that occurs immediately after the second activation, wherein the feature vector represents an intent of the particular customer regarding a particular customer event;
a plurality of dense neural network elements arranged in series, wherein the plurality of dense neural network elements is operably configured on the at least one computing processor to receive the feature vector and non-sequential inputs regarding a plurality of other customer interactions of the particular customer with the client, wherein the non-sequential inputs comprise a total number of orders, an order frequency, and an average order value, wherein each of the plurality of dense neural network elements is configured to consecutively apply: a first dense neural network, an first activation, a second dense neural network, a second activation, and a skip connection;
an output generator operably configured on the at least one computing processor to receive the output from the plurality of dense neural network elements and to generate: a distribution of times over which a particular customer event for the particular customer will occur or a likelihood estimation that the particular customer event for the particular customer will occur within a particular time period; and
a targeted marketing campaign generator configured to send messages or offers to the particular customer based on the distribution of times or the likelihood estimation.
2. The system of claim 1 , wherein each of the plurality of CDC elements is further configured to apply a first batch normalization between the first causal dilated convolution and the first activation.
3. The system of claim 1 , wherein each of the plurality of CDC elements is further configured to apply a second batch normalization between the second causal dilated convolution and the second activation.
4. The system of claim 1 , wherein each of the plurality of dense neural network elements is further configured to apply a first batch normalization between the first dense neural network and the first activation.
5. The system of claim 1 , wherein each of the plurality of dense neural network elements is further configured to apply a second batch normalization between the second dense neural network and the second activation.
6. The system of claim 1 , wherein each of the plurality of CDC elements is further configured to apply a first dropout between the first activation and the second causal dilated convolution.
7. The system of claim 1 , wherein each of the plurality of CDC elements is further configured to apply a second dropout between the second activation and the skip connection.
8. The system of claim 1 , wherein each of the plurality of dense neural network elements is further configured to apply a first dropout between the first activation and the second dense neural network.
9. The system of claim 1 , wherein each of the plurality of dense neural network elements is further configured to apply a second dropout between the second activation and the skip connection.
10. A system for predicting customer behavior, the system comprising:
at least one computing processor;
an embedding component operably configured on the at least one computing processor to receive and embed sequential inputs regarding a plurality of customer interactions of a particular customer with an online presence of a client, wherein the plurality of customer interactions comprise event type data, product data, and source data, wherein each of the plurality of customer interactions further comprise a timing of that interaction;
a plurality of causal dilated convolutional “CDC” elements arranged in series, wherein the plurality of CDC elements is operably configured on the at least one computing processor to receive the embedded sequential inputs, wherein the plurality of CDC elements is configured to output a feature vector, wherein each CDC element comprises two causal dilated convolutions with regularization that are bypassed with a skip connection, wherein the skip connection occurs immediately before the regularization, wherein the feature vector represents an intent of the particular customer regarding a particular customer event;
a plurality of dense neural network elements arranged in series, wherein the plurality of dense neural network elements is operably configured on the at least one computing processor to receive the feature vector and non-sequential inputs regarding a plurality of other customer interactions of the particular customer with the client, wherein the non-sequential inputs comprise a total number of orders, an order frequency, and an average order value, wherein each of the plurality of dense neural network elements comprises two dense neural networks with regularization that are bypassed with a skip connection;
an output generator operably configured on the at least one computing processor to receive the output from the plurality of dense neural network elements and to generate: a distribution of times over which a particular customer event for the particular customer will occur or a likelihood estimation that the particular customer event for the particular customer will occur within a particular time period; and
a targeted marketing campaign generator configured to send messages or offers to the particular customer based on the distribution of times or the likelihood estimation.