System and method for dimensionality reduction of vendor co-occurrence observations for improved transaction categorization

A method performed by one or more processors of an electronic bookkeeping system, the method comprising:

• generating, for each of a plurality of merchants known to the electronic bookkeeping system, a respective merchant characteristics vector including a plurality of first data fields, wherein each first data field of the plurality of first data fields corresponds to a different respective merchant of the plurality of merchants, and wherein the merchant characteristics vectors are associated with a first vector space;
• generating, for each merchant characteristics vector, a respective compressed merchant vector by compressing the merchant characteristics vector to include a reduced number of first data fields, wherein the reduced number of first data fields is selected based at least on a number of first data fields in the plurality of first data fields;
• generating, for each account of a plurality of accounts in the electronic bookkeeping system, a respective account characteristics vector including a plurality of second data fields having a same number of second data fields as the number of first data fields in the plurality of first data fields, wherein an n-th second data field of the plurality of second data fields corresponds to a same merchant of the plurality of merchants as an n-th first data field in the plurality of first data fields, and wherein each account is associated with a respective user of a plurality of users of the electronic bookkeeping system, and wherein the account characteristics vectors are associated with the first vector space;
• generating, for each account characteristics vector, a respective compressed account vector by compressing the account characteristics vector to include a reduced number of second data fields matching the reduced number of first data fields;
• embedding the compressed merchant vectors and the compressed account vectors into a compressed vector space having a reduced dimensionality corresponding to a dimensionality of the compressed merchant vectors and the compressed account vectors;
• and
• automatically sorting a set of new financial transactions associated with the plurality of users of the electronic bookkeeping system into the plurality of accounts based on the compressed merchant vectors and the compressed account vectors in the compressed vector space.