Heuristic customer clustering
Methods and apparatus are disclosed regarding an e-commerce system that places customers into a plurality of clusters and tailors services provided to a customer based on the cluster in which the customer is placed. In one embodiment, the e-commerce system defines the clusters based on purchase history data for customers having sufficient purchase history data. The e-commerce system then places customers without sufficient purchase history data into one of the defined clusters based on demographic data for the customer and demographic data for the customers in the cluster.
1. A method for clustering customers using a computing device having limited memory capacity, the method comprising:
splitting the customers into a first group comprising first customers who have conducted greater than a predetermined number of transactions and a second group comprising second customers who have not conducted greater than the predetermined number of transactions;
generating, based on transaction data of the first customers, a matrix that relates first customers to purchased products;
applying principal component analysis to the matrix to obtain a reduced matrix for the first customers;
generating, in accordance with a clustering large applications (CLARA) algorithm, a plurality of sample sets of the reduced matrix;
applying a partitioning around medoids (PAM) clustering algorithm to each of the plurality of sample sets to obtain a plurality of medoid sets comprising a medoid set per sample set;
clustering the first customers into a plurality of clusters based upon a medoid set of the plurality of medoid sets; and
placing each second customer into a cluster of the plurality of clusters based on demographic data for the respective second customer and demographic data for the first customers placed in each respective cluster.
2. The method of claim 1 , further comprising validating the plurality of clusters prior to placing each second customer into a cluster.
3. The method of claim 1 , further comprising recalibrating a quantity of dimensions of the reduced matrix until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
4. The method of claim 1 , further comprising recalibrating a quantity of clusters used by the clustering until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
5. The method of claim 1 , further comprising:
generating silhouette numbers for the first customers; and
determining, based on the silhouette numbers for the first customers, that the plurality of clusters are a valid clustering of the first customers.
6. The method of claim 1 , further comprising:
calculating, for each medoid set, a sum of distances to a nearest medoid of the medoid set; and
defining the clusters of the plurality of clusters based on the medoid set that has a smallest sum of distances.
7. The method of claim 1 , further comprising:
calculating, for each medoid set, a sum of Euclidean distances to a nearest medoid of the medoid set; and
defining the plurality of clusters based on the medoid set that has a smallest sum of Euclidean distances.
8. A non-transitory computer-readable medium, comprising a plurality of instructions, that in response to being executed, result in a computing device:
splitting customers into a first group comprising first customers who have conducted greater than a predetermined number of transactions and a second group comprising second customers who have not conducted greater than the predetermined number of transactions;
generating, based on transaction data of the first customers, a matrix that relates the first customers to purchased products;
applying principal component analysis to the matrix to obtain a reduced matrix for the first customers;
generating, in accordance with a clustering large applications (CLARA) algorithm, a plurality of sample sets of the reduced matrix;
applying a partitioning around medoids (PAM) clustering algorithm to each of the plurality of sample sets to obtain a plurality of medoid sets comprising a medoid set per sample set;
clustering the first customers into a plurality of clusters based upon a medoid set of the plurality of medoid sets; and
placing each second customer into a cluster of the plurality of clusters based on demographic data for the respective second customer and demographic data for the first customers placed in the respective cluster.
9. The non-transitory computer-readable medium of claim 8 , further comprising instructions that result in the computing device validating the plurality of clusters prior to the placing each second customer into a cluster.
10. The non-transitory computer-readable medium of claim 8 , further comprising instructions that result in the computing device recalibrating a quantity of dimensions of the reduced matrix until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
11. The non-transitory computer-readable medium of claim 8 , further comprising instructions that result in the computing device recalibrating a quantity of clusters used by the clustering until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
12. The non-transitory computer-readable medium of claim 8 , further comprising instructions that result in the computing device:
generating silhouette numbers for the first customers; and
determining, based on the silhouette numbers for the first customers, that the plurality of clusters are a valid clustering of the first customers.
13. The non-transitory computer-readable medium of claim 8 , further comprising instructions that result in the computing device:
calculating, for each medoid set, a sum of distances to a nearest medoid of the medoid set; and
defining the plurality of clusters based on the medoid set that has a smallest sum of distances.
14. A computing device, comprising
an electronic database comprising demographic data and transaction data for customers; and
a processor configured to:
split the customers into a first group comprising first customers who have conducted greater than a predetermined number of transactions and a second group comprising second customers who have not conducted greater than the predetermined number of transactions;
generate, based on the transaction data of the first customers, a matrix that relates the first customers to purchased products;
apply principal component analysis to the matrix to obtain a reduced matrix for the first customers;
generate, in accordance with a clustering large applications (CLARA) algorithm, a plurality of sample sets of the reduced matrix;
apply a partitioning around medoids (PAM) clustering algorithm to each of the plurality of sample sets to obtain a plurality of medoid sets comprising a medoid set per sample set;
cluster the first customers into a plurality of clusters based upon a medoid set of the plurality of medoid sets; and
place each second customer into a cluster of the plurality of clusters based demographic data for the respective second customer and demographic data for the first customers placed in the respective cluster.
15. The computing device of claim 14 , wherein the processor is further configured to validate the plurality of clusters prior to placing each second customer into a cluster.
16. The computing device of claim 14 , wherein the processor is further configured to recalibrate a quantity of dimensions of the reduced matrix until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
17. The computing device of claim 14 , wherein the processor is further configured to recalibrate a quantity of clusters used by the clustering until cluster validation determines the plurality of clusters are a valid clustering of the first customers.
18. The computing device of claim 14 , wherein the processor is further configured to:
generate silhouette numbers for the first customers; and
determine, based on the silhouette numbers for the first customers, that the plurality of clusters are a valid clustering of the first customers.
19. The computing device of claim 14 , wherein the processor is further configured to:
calculate, for each medoid set, a sum of distances to a nearest medoid of the medoid set; and
define the clusters of the plurality of clusters based on the medoid set that has a smallest sum of distances.
20. The computing device of claim 14 , wherein the processor is further configured to:
calculate, for each medoid set, a sum of Euclidean distances to a nearest medoid of the medoid set; and
define the plurality of clusters based on the medoid set that has a smallest sum of Euclidean distances.