chenglongma/librecapp

An application based on LibRec

README

Introduction - lib-rec-app.jar

This program is customized by Command Line (cli) and Configuration files.

Command Line

The following command line parameters are available:

[-cv <arg>] [-d <arg>] [-r <arg>] [-s]

More precisely,

1. -h : print help information.
2. -cv : The number of folds for cross validation, default: 5.
3. -d or --dataset: The dataset(s) to be processed, valid options are based on configuration file "default.properties".
4. -r or --recommender : The recommender(s) to be applied, valid options are based on configuration file "default.properties".
5. -s or --split : Whether to re-split the dataset, default: false.

For example, if you'd like to test user-based kNN and item-based kNN algorithms on datasets ml-100k and ml-1m, the folds of Cross Validation is 6, the command would be:

java -jar lib-rec-app.jar -d ml-100k ml-1m -r userknn itemknn -cv 6

* NB: The abbreviations of algorithms are based on Librec project, available at: AlgorithmList.

Configuration Files

The detailed configurations of this program and specific parameters of each algorithms are determined by various configuration files:

1. The default configuration
   1. Specify the default settings such as the data directories, dataset path.
   2. For more details, please refer to default.properties.
2. The dataset information
   1. Specify the input path, data format, etc.
   2. For more details, please refer to Dataset folder.
   3. NB: Will override same settings of default.properties.
3. The Recommender algorithms parameters
   1. Specify parameters of each algorithms.
   2. For more details, please refer to Algorithms folder.
   3. NB: Will override same settings of files in Dataset folder.
   4. Some parameter templates are available at: Algorithms Configuration.

Introduction - main.py

1. Open Anaconda3 -->Anaconda Prompt

   

2. Execute the command:

   conda activate py36

   The cli (Command Line Interface) would show:

   

3. Use cd command to change the current working directory to the folder of main.py

   cd C:\Users\hongl\OneDrive\Documents\LibRecApp

   The cli would show:

   

4. Execute the Python script to generate the merged files:

   python main.py -d [your dataset(s)] -r [your recommender(s)] -cv [the number of cross validation]

   Example:

   Experiment Condition:

   1. Dataset: ml-1m
   2. Recommender: userknn、itemknn
   3. CV: 5

   python main.py -d ml-1m -r userknn itemknn -cv 5

   The cli would show:

   

   Note:

   To get simple prompt usage:

   python main.py -h

   The cli would show:

   

5. Output file name:

   .../result/<dataset>/merge_[<recommender name>]/cv_<cv>_<timestamp>.csv

Some Tips of UNIX commands

A very useful simple-doc for UNIX introduction is available at: UNIX Survival Guide. (The cheat sheet at Page 30)

Please read carefully from Page 10 to Page 16. (Very easy to understand, I promise :)

"Screen" Command

Some times you may want to do something else and leave the server running at background. The screen command would be very useful.

Here is a simple introduction of screen.

You may use:

screen -ls

To list all sessions (attached and detached).

Example:

screen -r <Session name> (for example: screen -r 47647)

To resume (move to) certain screen. For convenience, you can type screen -r 4 as only one session name starts with "4". If you want to resume session "31185", you can type screen -r 311 at least, as there is another session named "31069".

Ctrl-A + Ctrl-D

To detach the current and return to parent session.

Example:

exit

Close current session, can be used on main session and any sub sessions.