A targeted association rule mining tool for extremely infrequent rules.
This tool can detect association rules with a very low minimum support number threshold by separating the antecedent item set and the consequent item set. Lower bound of the threshold is typically less than 10, however, it varies depending on the size and characteristics of the input data set.
To build the tool, JDK 11 or higher and Apache Maven is required.
Usage: java -jar cTARM-X.Y.jar MIN_SUPPORT MIN_CONFIDENCE PATH_TO_TSV_DATA REGEXP_OF_CONSEQUENT_ITEM
% git clone https://github.com/yhayase/cTARM.git
...
% cd cTARM
% mvn package
...
[INFO] BUILD SUCCESS
...
% java -jar target/cTARM-X.Y.jar 1 0.1 data/small-input-example.tsv '^x'
15 rules are found.
Rule: [a] => [xA, xB] w/ freq=2 conf=0.666667
Rule: [a] => [xA] w/ freq=3 conf=1.000000
Rule: [b, a] => [xA, xB] w/ freq=2 conf=1.000000
Rule: [c, b, a] => [xB, xA] w/ freq=1 conf=1.000000
Rule: [c, a] => [xA, xB] w/ freq=1 conf=1.000000
Rule: [b] => [xB, xA] w/ freq=2 conf=1.000000
Rule: [b, c] => [xB, xA] w/ freq=1 conf=1.000000
Rule: [c] => [xB, xA] w/ freq=1 conf=1.000000
Rule: [d] => [xC] w/ freq=1 conf=1.000000
50 [msec] (for 15 rules)