rmkujala / dmproj Goto Github PK

Project repository for the Algorithmic methods for data mining course project.

Group Members:
Rainer Kujala, [email protected]
Olli-Pekka Koistinen, [email protected]


Code usage:

To obtain the figures shown in the report, the following steps need to be performed:

0. Set up the param data_dir (containing the orig. data set and future results) in settings.py
1. Run python prepro.py
	- This the plots in the report and computes the preprocessed data sets
2. Run brute_force.py with all input parameter combinations [0-2, 0-7]:
	pypy brute_force.py 0 0
	pypy brute_force.py 0 1
	...
	pypy brute_force.py 2 7
	(this takes a while)
3. Run plot_brute_force.py (using python) for plotting the results for brute force.
4. Run sh create_identity_indices.sh
5. Run sh create_speedup_indices.sh
6. Run speedups.py (using pypy) with all input parameter combinations as in step 2.
7. Run plot_speedups.py (using python) for plotting the speedup curves
8. Run approx.py (using pypy) with input parameter combinations [0-2, 0-7, 0-1]
	(the last one is for selecting the approximation threshold (1.5 or 1.2))
9. Run plot_approx.py (using python) to plot the speedups and mean estimation accuracies.

10. To assert that the results are valid, run assert_results.py (with python)