A Unified Influence Maximization Processing Framework for Independent Cascade Model and Its Extensions

The first line of a graph data file consists of two integers; the first one is the number of nodes and the second one is the number of edges.

The second to last line also consist of two integers which represent node-id. node-id should range over from \(0\) to \(\text{<number of nodes>} - 1\). When the graph is directed, an edge starts from the first node-id to the second id. Otherwise, the order of two integers is not important.

The following is an example of a data file that represent the above graph.

6    5    # <# of nodes>  <# of edges>
0    1    # edge description
1    2
2    3
3    1
4    5

Here we refer to two publicly available dataset repositories.

• Wei Chen's project page
  • NetHEPT and NetPHY
  • DBLP

  The data format used here complies with our data file format specification. After downloading them, you can apply them to our application without any modification.

• Stanford Large Network Dataset Collection
  • Epinion
  • Stanford
  • Patent
  • LiveJournal
  • and more …

  The data format used here does not follow our data file format specification. Thus, we provide a windows data preprocessor binary in Download Application Binary. The usage is that after (1) putting the preprocessor in dataset location and (2) moving into the dataset location in window command line, execute the following command.

  path-to-preprocessor> IPA_preprocessor.exe <datafile>
  

  After executing, in dataset location, the preprocessed datafile named preprocessed_<datafile> and id-matching table from original dataset to preprocessed dataset matching_table_<datafile> are generated.

• Network
  • path : the path of network dataset file
  • directed : when checked, the network is directed, otherwise, undirected
• Weighting Model
  • One of IC and WC
    • IC : an uniform probability is assigned to all edges
    • WC : for a node \(v\) and \(v\)'s in-neighbor set \(N_{in}(v)\), \(\frac{1}{|N_{in}(v)|}\) is assigned to each edge \(u \in N_{in}(v)\)
  • PP : the uniform propagation probability when IC is checked

• Diffusion Model
  • One of IC, t-IC, IC-N, and t-IC-N
• General Parameter
  • k : the number of seed nodes
  • Threshold : the lower bound of the probability that an influence path can have
  • Multiplier : the number of entries in priority queue (Multiplier \(\times\) k)
• t- options
  • T : the ending stage of influence diffusion
  • alpha : decaying speed of \(\delta(\cdot)\), the larger value means the higher decaying speed
• -N option
  • quality factor : the probability that a positive opinion is kept when a node is activated

• download the desktop application binary(ipa-desktop.zip) and data converter(ipa-preprocessor.zip – it's optional) from Download Application Binary
• download datafile (e.g. NetHEPT and NetPHY) from Available Datasets
• extract ipa-desktop.zip into C:/IPA Demo/Desktop App

  

• extract graphdata.zip from Wei Chen's project page
• extract ipa-preprocessor.zip into C:/IPA Demo/Data (optional to handle dataset from Stanford Large Network Dataset Collection)

  

• move to C:/IPA Demo/Desktop APP
• execute IPA_Demo.exe
• set dataset

  

  

• set parameters and push submit button
  • default value is enough to see result

  

It takes a long time to get influence cloud. Thus, we recommend to use this feature only for smaller dataset.

After getting seed nodes from IPA algorithm,

• push draw button, a new pop-up window appears

  

  

• to draw again, you should run IPA again – push submit button again