In this module, we will now look at customer satisfaction influence analysis by using this tool of network analysis. This is a bit of an intro to network analysis, some basics. First of all, when we're engaging in network analysis, what types of data should we analyze? Well, the list is not that long, at least here. I suggest social media data, because if we're trying to understand how people are interacting with other people and forming networks, then really social networks are perfect dataset to analyze this phenomenon. But there's, of course, more datasets, I'm just suggesting that those datasets are probably not easily accessible publicly. Your company might have datasets of how people interact with one another regarding customer satisfaction. But for this course, we will look primarily or solely at social media data. Now, a definition or a conceptual understanding of network analysis can be found here. I'm not going to read it all, but just the first few sentences. The basic idea of a social network is very simple. A social network is a set of actors that may have relationships with one another. Now, in network analysis, usually these networks are visualized. This is a visual that signifies a very simple network. Now there's various terms used in network analysis that signify the same concepts, and so I'll try to use as many of them as possible just so that you know all of them or at least many of them that are used commonly. This is typically called either a network, or a graph, or they combine the words and they say network graph. Network graphs have components, these components are, first, nodes or vertexes, and again, the same concept, different terms. Joe is either a node or Joe is a vertex. In this case, we're using nodes as people, but nodes can be anything, nodes can be organizations, nodes can be animals, anything that can interact with something else. Nodes can be bacteria, so on and so forth. Nodes are connected via ties or edges or links. Again, same concept, different terms. Now, you can define what these ties represent, what these edges represent. Let's define these edges as friendships. This is a friendship network. Joe is friends with Bob, Bob is friends with Ann, Ann is friends with Joe. Now, edges can either be weighted or unweighted, and if they're weighted, this is called a weighted network graph or a weighted network or a weighted graph, again, the combinations are bound. But what does weighting signify? If the edge signifies a friendship, then the weighting signifies the intensity of that friendship. In this case, Joe is best friends with Bob. Bob is maybe an acquaintance with Ann, and Ann is friends with Joe but not best friends. This is what this weighted network graph signifies. Graphs can also be directed. What does that mean? Well, if this is a friendship graph, this is sad. Because what this means is Joe believes that Bob is a friend and Bob believes that Joe is a friend, it's a mutual friendship. But then in the case of Bob and Ann, Bob thinks Ann is a friend, but Ann does not think Bob is a friend. In the same way, Ann thinks Joe is a friend, but Joe does not think Ann is a friend. Now this is a sad graph, but it's a good example of a directed graph. Graphs can have direction. Then we can combine all of this and do a weighted and directed network graph where Joe and Bob consider mutually each other to be best friends. Bob considers Ann to be an acquaintance, Ann does not consider Bob to be an acquaintance. Whereas Ann considers Joe to be a friend, but Joe does not consider Ann to be a friend. You can see you can play around with these components to represent different relationships between entities, between nodes or vertexes using edges and these edges can be weighted and they can be directed. You can represent a network graph as well from a visual as well as writing down the network graph as an edge list. What you see here is a directed network graph on the left and an edge list that represents that directed network graph. You see Joe, Bob, and in the next row you see Bob, Joe, why? Joe and Bob listed twice. Well, it's signifying in a directed graph that there's a direction going from Joe to Bob, but then there's also a direction of an edge going from Bob to Joe. Then we see Bob, Ann, but we don't see Ann, Bob, because again remember Ann doesn't consider Bob a friend. Then we see Ann, Joe but not Joe, Ann because remember Joe does not consider Ann to be a friend. You can represent this visual on the left via this edge list on the right. There's also names for special networks. Now I don't know if this is special, but in the case of certain phenomenon, it could be special where you have a singleton, which you might have a point or a vertexes a node that is just by itself, it's not connected to anybody else. That might be an interesting phenomenon depending on what you're studying. That's usually called a singleton amidst all other nodes and vertexes. You might also have what is called the click. We've heard that term probably before in society. But in network analysis it's not necessarily an exclusive group of people, it's just an interesting group of nodes, and all these nodes must be connected. Let's say Joe, Bob, and Ann for whatever reason in enlarge or picture that Joe, Bob, and Ann form some special function, so they're interesting, and so they would be called a click. Then we'll see clusters. Let's pretend Joe, Bob, and Ann are just part of this larger graph, this larger network. You see that this larger network even visually there seems to be two parts to this larger network, and those are called clusters. They're group of nodes that have many connections and it's clearly more connected to one another than the graph is as a whole are compared to other sub-graphs. Clusters are very important phenomenon within network analysis because you'll see separation between groups of people for whatever reason, and that might be interesting to what you're studying. These are just a few things that form a basic foundation for network analysis.
