Back to results
US 11,237,938 B2Granted
Click heatmap abnormality detection method and apparatus
AssigneeBeijing Gridsum Technology Co., Ltd.
InventorZhenhua Liu
Life of this patent
Estimating the term…
Abstract
A click heatmap abnormality detection method and apparatus, comprising: obtaining a first click heatmap, and dividing the first click heatmap into a plurality of regions (S100); for each region among the plurality of regions, respectively determining a click probability of a click action corresponding to each click source happening in the region (S200); for each click source, determining an abnormal click region corresponding to the click source according to the click probabilities (S300). The described solution determines abnormal click regions according to the click probabilities of the click actions corresponding to the click sources happening in the regions, without the need for manual identification, and with high accuracy and identification efficiency.
Claims12
A method for detecting abnormality in a click heatmap, comprising:
- capturing a click action within a web page to generate a first click heatmap;
- dividing the first click heatmap into a plurality of regions, comprising:
- dividing the first click heatmap evenly into rectangular regions, wherein a quantity of the rectangular regions is row×col, row is a quantity of rows of the rectangular regions in the first click heatmap, and col is a quantity of columns of the rectangular regions in the first click heatmap;
- determining, for each of the plurality of regions, a click probability of a click action corresponding to each click source occurring in said region;
- determining, for each click source, an abnormal click region corresponding to said click source according to the click probability, wherein the abnormal click region is at least one of the plurality of regions;
- and
- determining abnormal traffic according to the abnormal click region;
- wherein determining, for each click source, an abnormal click region corresponding to said click source according to the click probability comprises:
- for each click source P,
- determining first relative entropy D1, of said click source P relative to other click sources P, to be:
- D1=Σi=1,j=1row,coldi,jpi,j log(pi,j/pi,j);
- determining, under a condition that D1 is less than a first preset threshold, all of di js when Σdi,j is maximum;
- and
- determining one of the rectangular regions as the abnormal click region, wherein the one of the rectangular regions corresponds to di,j equal to 0 in the di js when Σdi,j is maximum;
- wherein i is a row number of the rectangular regions, j is a column number of the rectangular regions, di,j is a region-anomaly identifier, pi,j is the click probability of said click source P in a rectangular region at an i-th one of the rows and a j-th one of the columns, pi,j is the click probability of the other click sources P the rectangular region at the i-th one of the rows and the j-th one of the columns, and di,j is equal to 0 or 1.
12 more elements
A method for detecting abnormality in a click heatmap, comprising:
- capturing a click action within a web page to generate a first click heatmap;
- dividing the first click heatmap into a plurality of regions, comprising:
- dividing the first click heatmap evenly into rectangular regions, wherein a quantity of the rectangular regions is row×col, row is a quantity of rows of the rectangular regions in the first click heatmap, and col is a quantity of columns of the rectangular regions in the first click heatmap;
- determining, for each of the plurality of regions, a click probability of a click action corresponding to each click source occurring in said region;
- determining, for each click source, an abnormal click region corresponding to said click source according to the click probability, wherein the abnormal click region is at least one of the plurality of regions;
- and
- determining abnormal traffic according to the abnormal click region;
- wherein determining, for each click source, an abnormal click region corresponding to said click source according to the click probability comprises:
- determining di,js corresponding to each click source when H is minimum, wherein H is a sum of second relative entropy of each click source relative to other click sources, and
- H=ΣP∈U{Σi=1,j=1row,col[di,jpi,j log(pi,j/pi,j)+(1−di,j)pi,j log(pi,j/δb)]};
- determining, for each click source, one of the rectangular regions as the abnormal click region, wherein the one of the rectangular regions corresponds to di,j equal to 0 in the di,js corresponding to said click source;
- wherein i is a row number of the rectangular regions, j is a column number of the rectangular regions, di,j is a region-anomaly identifier, pi,j is the click probability of said click source P in a rectangular region at an i-th one of the rows and a j-th one of the columns, pi,j is the click probability of the other click sources P the rectangular region at the i-th one of the rows and the j-th one of the columns, di,j is equal to 0 or 1, δ is a penalty coefficient, b is a benchmark click probability, and b=1/(row×col).
9 more elements
A method for detecting abnormality in a click heatmap, comprising:
- capturing a click action within a web page to generate a first click heatmap;
- dividing the first click heatmap into a plurality of regions, comprising:
- dividing the first click heatmap evenly into rectangular regions, wherein a quantity of the rectangular regions is row×col, row is a quantity of rows of the rectangular regions in the first click heatmap, and col is a quantity of columns of the rectangular regions in the first click heatmap;
- determining, for each of the plurality of regions, a click probability of a click action corresponding to each click source occurring in said region;
- determining, for each click source, an abnormal click region corresponding to said click source according to the click probability, wherein the abnormal click region is at least one of the plurality of regions;
- and
- determining abnormal traffic according to the abnormal click region;
- wherein determining, for each click source, an abnormal click region corresponding to said click source according to the click probability comprises:
- determining, for each of the rectangular regions, a maximum among ratios of the click probability of every two click sources, corresponding to said rectangular region;
- and
- determining one of the rectangular regions as the abnormal click region, wherein the one of the rectangular regions corresponds to the maximum greater than a second preset threshold, the second preset threshold is greater than one, and the maximum is a ratio of the click probability, corresponding to the one of the rectangular regions, between said click source and another click source.
8 more elements
Dependencies are inferred from the claim text.