nju-websoft / QGSTP-HB

A Fast Hop-Biased Approximation Algorithm for the Quadratic Group Steiner Tree Problem

This is the source code of the paper 'A Fast Hop-Biased Approximation Algorithm for the Quadratic Group Steiner Tree Problem'.

Table of contents

• Directory Structure
• Environment
• Data
  • Hub Label Construction
• Run the HB Algorithm
  • Experiment 1 (Table 2)
  • Experiment 2 (Figure 3 and Figure 4)
  • Experiment 3 (Table 3)
  • Experiment 4 (Table 4)
  • Extended Experiment 1 (Figure 5)
  • Extended Experiment 4 (Table 6 and Table 7)
• Other Algorithms
• License
• Citation

Directory Structure

Directory /src/main/java contains all the source code based on JDK 11.

• Directory /src/main/java/graphtheory contains our implementation of the HB algorithm

• Directory /src/main/java/driver/data includes some classes for generating graph data and queries

• Directory /src/main/java/driver/work conducts the experiments for the HB algorithm

• Directory /src/main/java/mytools consists of some tools

• Directory /src/main/resources provides a sample configuration to run this project

Environment

• MySQL
• JDK11
• Maven
• 180G Memory

Data

Our dataset is available on .

Import the data to your MySQL database.

Hub Label Construction

Unfortunately, due to the limit of space, we could not directly provide the data of hub labeling for DBpedia and LUBM-250U. You might take a long time to construct the hub labeling. Otherwise, if you don't want to do that, you may choose to run the simple script (marked by _simple) to run the algorithm without DBpedia and LUBM-250U.

First set the configuration /src/main/resources/config.properties as follows:

• IP, PORT : the IP and the port to connect to your MySQL database,
• USER,PASS : the username and the password to log in your MySQL database.

Then run the following script (it might cost a very long time):

bash Construct_Hub_Label.sh

Run the HB Algorithm

First set the configuration /src/main/resources/config.properties as follows:

• IP, PORT : the IP and the port to connect to your MySQL database,
• USER,PASS : the username and the password to log in your MySQL database.

Experiment 1 (Table 2)

Run the following script:

bash Experiment_for_Table_2.sh

For each graph, the result for each query is record in file Table_2_[graph].log. For example, Table_2_mondial.log contains the results for all 39 queries, include $\alpha=0.1$, $\alpha=0.5$ and $\alpha=0.9$. And we can find the query result for query 1 like:

Expr 1:
Generating keyword...
Running algorithm...
Finish!
The number of graph nodes is 207631, the number of graph edges is 811556.
The number of keyword is 2.
The size of the smallest group is 100.
Algorithm: HB
alpha= 0.1
Runtime: 0.169 s.
Complete! cost(T)=0.7636719283306959.

After obtaining the runtime for each query, we can calculate the average runtime in Table 2.

Experiment 2 (Figure 3 and Figure 4)

Run the following script:

bash Experiment_for_Figure_3_and_4.sh

Then we get the similar result Figure_3_and_4_[graph].log. After obtaining the runtime, the number of groups (g) and the size of the smallest group (f), we can draw Figure 3 and 4.

Experiment 3 (Table 3)

Run the following script:

bash Experiment_for_Table_3.sh

Then we get the similar result Table_3_[graph].log. We can calculate empirical approximation ratio in Table 3 by combining the cost of the answer to each query, and the optimum answer (computed by B3F additionally).

Experiment 4 (Table 4)

Run the following scripts:

bash Experiment_for_Table_4.sh

The result is recorded in Table_4_woPruning_[graph].log, Table_4_woPP_[graph].log, Table_4_woPI_[graph].log. After obtaining the runtime for each query under each setting, we can calculate the average runtime in Table 4.

Extended Experiment 1 (Figure 5)

We can obtain the figure by reusing Table_2_[graph].log generated by Experiment_for_Table_2.sh.

Extended Experiment 4 (Table 6 and Table 7)

Run the following scripts:

bash Experiment_for_Table_6_and_7.sh

The result for Table 6 is recorded in Table_6_woPruning_[graph].log and the result of Table 7 is recorded in Table_7_woPP_[graph].log and Table_7_woPI_[graph].log. After obtaining the runtime for each query under each setting, we can calculate the average runtime.

Other Algorithms

Please refer to the following repositories if you want to run other algorithms in our experiments:

• the EO algorithm: QGSTP,
• the B3F algorithm: B3F.

License

This project is licensed under the apache 2.0 License - see the LICENSE file for details

Citation

If you think our algorithms or our experimental results are useful, please kindly cite our paper.