To run the program with dataset provided and default values for _min_support_(0.15) and *min_confidence*(0.6)

python apriori.py -f INTEGRATED-DATASET.csv

To run program with dataset

python apriori.py -f INTEGRATED-DATASET.csv -s 0.17 -c 0.68

Best results are obtained for the following values of support and confidence:

Support : Between 0.1 and 0.2

Confidence : Between 0.5 and 0.7

The primary entrypoint to the package is the run_apriori() method. You will need to load your data into an iterable object, either using the data_from_file() function or the FileIterator() class, and then pass this to run_apriori() along with a minimum support and a minimum confidence. The data_from_file() function loads all the data into memory at once, which is fast for smaller files but can cause problems for larger ones. The FileIterator() class keeps the data on disk and loads only what is needed as it is needed, which is slow for smaller data sets but avoids memory issues that can slow the process down for larger ones.

If the records you are processing is columnar, where the changing the order of the values in a row affects its meaning, be sure to set the ordered flag when loading your data with data_from_file() or FileIterator(). This will convert each value in the row to an (index, value) pair so the algorithm treats the values as non-reorderable. By default, the algorithm assumes the order of the values in a row is inconsequential.

run_apriori() returns a pair, (itemsets, rules). The write_itemsets() and write_rules() functions are provided as a convenience for saving the results to file.

Example Usage:

from apriori import run_apriori, data_from_file, FileIterator

data_csv_path = 'your_source_data.csv' itemsets_csv_path = 'itemsets_save_loc.csv' rules_csv_path = 'rules_save_loc.csv'

dataset_is_small = True # This depends on your hardware data_is_columnar = True # Whether the order of the values appearing in a given row matters

min_support = .15 min_confidence = .6

if dataset_is_small:

data_iterable = data_from_file(data_csv_path, ordered=data_is_columnar)

else:

data_iterable = FileIterator(data_csv_path, ordered=data_is_columnar)

itemsets, rules = run_apriori(data_iterable, min_support, min_confidence)

write_itemsets(itemsets_csv_path, itemsets, ordered=data_is_columnar) write_rules(rules_csv_path, rules, ordered=data_is_columnar)

MIT-License

INTEGRATED-DATASET.csv is a copy of the “Online directory of certified businesses with a detailed profile” file from the Small Business Services (SBS) dataset in the NYC Open Data Sets