Differences

This shows you the differences between two versions of the page.

Link to this comparison view

online:snminer [2009/05/26 10:45] (current)
Line 1: Line 1:
 +====== Social Network Miner ======
 +
 +This plugin reads a process log and generates social networks that can be used as a starting point for SNA. We can apply several techniques to analyze the social networks, e.g., find interaction patterns, evaluate the role of an individual in an organization,​ etc. 
 +
 +===== Prerequesites =====
 +
 +  - Open an event log file in ProM (The log should have originators)
 +  - Choose Social Network Miner from the Mining menu
 +
 +===== Settings =====
 +
 +{{documentation:​socialnetwork:​sna2.gif|}}\\
 +
 +**Figure 1.** Social Network Miner settings allow for the selection of the metrics to be calculated
 +
 +The plugin provides five kinds of metrics to generate social networks. They are "​handover of work", "​subcontracting",​ "​working together",​ "​similar task", and "​reassignment"​. ​
 +
 +  * Handover of work metric: Within a case (i.e., process instance) there is a handover of work from individual //i// to individual //j// if there are two subsequent activities where the first is completed by //i// and the second by //j//. This notion can be refined in various ways. For example, knowledge of the process structure can be used to detect whether there is really a causal dependency between both activities. It is also possible to not only consider direct succession but also indirect succession using a "​causality fall factor"​ //beta//, i.e., if there are 3 activities in-between an activity completed by //i// and an activity completed by //j//, the causality fall factor is //​beta^3//​. ​
 +  * Subcontracting metric: The main idea is to count the number of times individual //j// executed an activity in-between two activities executed by individual //i//. This may indicate that work was subcontracted from //i// to //j//. All kinds of refinements mentioned in //Handover of work metric// are also possible.
 +  * Working together metric: This ignores causal dependencies but simply counts how frequently two individuals are performing activities for the same case. If individuals work together on cases, they will have a stronger relation than individuals rarely working together. There are three kinds of methods to calcuate working together metric. The first one is dividing the number of joint cases by the number of cases in which individual //i// appeared. It is important to use a relative notation. For example, suppose that individual //i// participates in three cases, individual //j// participates in six cases, and they work together three times. In this situation, //i// always work together with //j//, but //j// does not. Thus, the value for //i// to //j// has to be larger than the value for //j// to //i//. Alternative metrics can be composed by taking the distance between activities into account.
 +  * Similar task metric: It does not consider how individuals work together on shared cases but focuses on the activities they do. The assumption here is that people doing similar things have stronger relations than people doing completely different things. Each individual has a "​profile"​ based on how frequent they conduct specific activities. There are many ways to measure the "​distance"​ between two profiles thus enabling many metrics. There are four kinds of distance metrics. //Euclidean distance// is the "​ordinary"​ distance between two points that one would measure with a ruler. (It only gives good results if performers execute comparable volumes of work.) //​Pearson’s correlation coefficient//​ is frequently used to find the relationship among cases. //​Similarity coefficient//​ is a statistic used for comparing the similarity and diversity of sample sets. //Hamming distance// does not consider the absolute frequency but only whether it is 0 or not.
 +  * Reassignment metric: It considers the type of event. Thus far we assumed that events correspond to the execution of activities. However, there are also events like reassigning an activity from one individual to another. For example, if //i// frequently delegates work to //j// but not vice versa it is likely that //i// is in a hierarchical relation with //j//. From a SNA point of view these observations are particularly interesting since they represent explicit power relations.
 +=====  =====
 +
 +The bottom of Figure1 shows the filtering option. Users can also filter originators according to their frequencies. Users also specify an organizational model file for the process log. It can be used to analyze social network using Social network analysis plug-in. ​
 +
 +===== Results =====
 +
 +{{documentation:​socialnetwork:​sna-result.gif|}}\\
 +
 +**Figure 2.** Social Network Mining Result ​
 +
 +Figure2 shows an example of the mining result. The generated social network is displayed as a form of matrix (top right) and graph (bottom). One can remove low frequent arcs using the sliderbar. "​Remove isolated nodes" button enables users to remove the isolated nodes from the screen. ​
 +
 +===== Further steps =====
 +
 +From the result panel, users can execute //Social network analysis// plug-in to perform social network analysis. ​
 +
 +===== Publications =====
 +
 +{{page>​blogs:​pub2005:​discovering_social_networks_from_event_logs&​noeditbtn&​firstseconly}}
 +
 +{{page>​blogs:​pub2007:​towards_comprehensive_support_for_organizational_mining&​noeditbtn&​firstseconly}}
 +