Interactive data exploration
Provided are methods and systems comprising receiving a data set comprising a plurality of records having a plurality of fields, displaying a plurality of graphical objects representing the plurality of records, receiving a selection of one or more of the plurality of fields, determining, based on the selected one or more of the plurality of fields, a similarity score between each of the records of the plurality of records, and adjusting, based on the similarity scores, the plurality of graphical objects.
1. A method comprising:
receiving a data set comprising a plurality of records comprising a plurality of fields, wherein each record of the plurality of records comprises a value for each field of the plurality of fields, and wherein each field of the plurality of fields comprises a field type;
displaying a plurality of graphical objects representing the plurality of records;
receiving a selection of one or more fields of the plurality of fields;
determining, for each pair of values of the plurality of records and for each field of the one or more fields, a vector of similarities based on the field type corresponding to each field of the one or more fields, wherein each of the vectors of similarities comprises a similarity score for a corresponding field of the one or more fields;
determining, based on the vectors of similarities, a combined score;
determining, based on the combined score and a multidimensional scaling technique, a display position for each record of the plurality of records, wherein the multidimensional scaling technique is applied based on the vectors of similarities; and
adjusting, based on the display position for each record of the plurality of records, the plurality of graphical objects;
wherein determining the vector of similarities based on the field type corresponding to each field of the one or more fields comprises, for each pair of values for each field (P) from two records (vi, vj) of the plurality of records, determining the vector of similarities;
wherein the field type comprises a numeric field, a categorical field, a date field, a time field, or a geolocation field; and
wherein determining the vector of similarities based on the field type corresponding to each field of the one or more fields comprises at least one of:
for numeric fields (nF) of the one or more fields for the two records (vi, vj), determining a numeric vector for each numeric field by determining 1—an absolute value of an average of values of each numeric field of the record vi—an average of values of each numeric field of the record vj/a range of the numeric fields (nF);
for categorical fields (cF) of the one or more fields for the two records (vi, vj), determining a categorical vector for each categorical field by determining a dot-product of a histogram of values of each categorical field of the record vi and a histogram of values of each categorical field of the record vj; and
for the one or more fields of the two records (vi, vj), determining a dot-product of the numeric vectors and the categorical vectors, resulting in the vector of similarities for each field of the one or more fields between the two records (vi, vj).
2. The method of claim 1 , wherein the multidimensional scaling technique comprises one or more of: metric multidimensional scaling or metric least squares scaling.
3. The method of claim 1 , wherein adjusting the plurality of graphical objects comprises: moving graphical objects with similar similarity scores closer together and moving graphical objects with dissimilar similarity scores further apart.