We all must have solved and learned graphs in school days and have a picture of a huge grid having x-axis and y-axis with points connected to each other and having distance as standard unit size. Here Graph theory is the study of graphs, which are mathematical structures used to model relations between objects and entities. A graph in this context is made up of vertices that are connected by edges.
An example of a graph.
Graph analytics grew upon the concept of finding relationships within data to yield more relevant and precise information. Graphs are ideal to store, connect, establish relations between diverse entities/objects, and make inferences from seemingly complex unrelated data. Some of the commonly used unsupervised machine learning graph analytics methods are community detection, label propagation, PageRank, centrality analysis, connectivity analysis, and path analysis.
Understanding graphs and inference:
A graph is supposed to be interpreted as per the use-case/domain to dwell maximum use case and applicability. Understanding the database and what the nodes and vertices indicated is the first step towards the interpretation of graphs.
Everyone is aware of google maps and use it on a regular basis. Google uses the concept of path analytics to find the most optimum or rather multiple similarly optimized pathways to reach your destination. Often multiple paths are available for a user to reach the desired destination. A path may be less congested but includes a detour while another path may be highly congested with less road distance. These alternatives are generated by assigning suitable weights to the nodes and edges assigned by Google for obtaining the most optimum possible paths. You can even add constraints to the path. Suppose you have to drop off your kid at school while going to work, you add a constraint of the location of the school (node J) occurring before you reach your office(node G). Then a path passing through these nodes is constructed considering the traffic, roadblock, construction, distance, etc. Dijkstra’s Algorithm is one of the most common algorithms used to calculate the distance between nodes. In essence, destinations and constraints in between are assigned as nodes, and roads are assigned as edges to obtain the path.
Example of graph network
Apps like Neo4j, Graphx Spark, Gephi are used for developing graphical models using numerical and categorical databases.
Application examples:
Fraudulent is unjustifiably claiming or being credited with some accomplishments. Many enterprises employ large teams of trained investigators to determine whether a transaction is likely to be fraudulent or not. Outliers in some models often point towards abnormal phenomenons. Tracing and finding how it is developed often leads to some new discoveries.
For eg:
1. Finding the number of incoming and outgoing edges from a node and tracing out which nodes are performing the maximum amount of outgoing activity can help in the identification of abnormal activities and in some cases lead to fraudulent activities.
2. Each day, millions of phone calls are made, but only a fraction of them are potential scams. Graph technology can explore relationships among callers, phone numbers, and when collaborating with ML it can develop trained models to detect which are scam calls.
3. Community analysis can be used to identify key persons in a community. This can be useful in political campaigns to identify how to treat commodities and take informed decisions.
4. Graph analytics can be used in electrical system design to identify which nodes are weak, strong, and how the interconnectivity is established. Protection of nodes by identifying which node is the most important and can lead to disaster if hampered.
5. In Natural Language Processing commonly known as NLP, a particular methodology exists which is known as Bag of words. The features of the words associated with a node along with vocabulary-based feature vectors can calculate quantitative values of the structural position of a node in a graph. Inference on the basis of degree centrality, closeness centrality, betweenness centrality is a very effective method.
Conclusion:
Advances in understanding and intermixing of diverse domains and functioning on collective intelligence is a possibility due to the emerging possibilities of overlapping domains. The amalgamation of various fields and domains is transpiring and is leading to the manifestation of human imagination. The linkage between medicine, technology, psychology, sociology, and various other fields seems to be able to get in sync with each other and we are thus able to understand abstract ideas and perspectives of existence with a scientific backup.
Comments