Blog 4 for Social Media Analytics

Hi, everyone. See you again. Time really flies. In the recent lectures, we learned about link analysis and three corresponding link analysis algorithms named respectively HITS(Hyperlink-induced Topic Search), PageRand, EigenRumor. Link analysis refers to analysis of relationships and information flow between individuals, groups, organizations, servers, and other connected entities. Not only do we know about their intrinsic values, but also we should understand clearly their positions in social network relations. Frequenlty, when searching a keywordon web page, we look forward to acquiring the most authoritative content. So it's necessary to use some solutions or approaches to process link analysis.

1. HITS
According to HITS algorithm, after the user enters the keyword, matching algorithm to return the page to calculate two values, one is the hub value, the other is the authoritative value, these two values are interdependent, influencing each other. A good hub represented a page that pointed to many other pages, and a good authority represented a page that was linked by many different hubs.
2. PageRank
PageRank is an algorithm used by Google Search to rank websites in their search engine results. PageRank is a way of measuring the importance of website pages. PageRank works by counting the number and quality of links to a page to determine a rough estimate of how important the website is. The underlying assumption is that more important websites are likely to receive more links from other websites.
3. EigenRumor
EigenRumor scores each blog entry by weighting the hub and authority scores of the bloggers based on eigenvector calculations. This algorithm enables a higher score to be assigned to the blog entries submitted by a good blogger but not yet linked to by any other blogs based on acceptance of the blogger's prior work.

In social environment, we are faced with the choice of information overload. In order to filter information which is useless for us, we need to use computer to process the huge amount of information. Therefore, recommender systems play an important role in recommending what users should choose. We should know clearly their own interests as well as value of the items or something similar to what users have consumed before. There are two methods to process content-based recommendation. One is similar based recommendation which calculate similarity between user profile and the items to match them. The other is supervised learning which needs negative samples to train a classifer.

In addition, collaborative filtering performs recommend based on the history of user-item interactions by analysing tastes and preferences of users. There are two types of collaborative filtering named neighbourhood-based methods and model-based mathematical models to represent users, items and their relations. Finally, they both generat a ranked list of items for an user.

Reference: 
[1] "Facts about Google and Competition". Archived from the original on 4 November 2011. Retrived 12 July 2014.
[2] The EigenRumor algorithm for ranking blogs(2005), by Ko Fujimura, Takafumi Inoue, Masayuki Sugisaki.

Blog 3 for Social Media Analytics

Hi, everyone. After studying the lecture 5 and lecture 6, I have something to share with you. In terms of social psychology, I think it contains a lot of abstract theories. Living in the social environment, our thoughts, feelings and behaviors are actually influenced by others since we were born. We can say that it’s our inborn instinct to be social. Whatever it’s right or wrong, we obtain information in life or in some social platforms and subconsciously we affected by them and diffuses relevant information to others who connected with us.

Figure 1 Social Psychology

On our own terms, we show our behavior with our internalized standards, because our performance is connected with what we think in our minds. However, our actions are judged by people around us. In order to present ourselves to others much better, we should have social comparison. If we do something wrong or disapproved by others, we must introspect ourselves and change as soon as possible. In social environment, after daily we contact with people for a long time, we can understand what they will do next when they confront us or different cases.

A famous theory called Social Cognitive Theory (SCT), holds that portions of an individual's knowledge acquisition can be directly related to observing others within the context of social interactions, experiences, and outside media influences. For example, when we were children ,our parents taught us to speak and we imitate their language contents. In social process, usually there will be new objects appearing constantly. Initially, people are unfamiliar with them, so maybe they are not accepted by people. After a period, maybe more people know about the potential information of new objects and accept them gradually.

In a social network, each participant share own unique difference to contribute to the group in which they affect each other. There are three levels of cognitive processing named cognition, metacognition, epistemic cognition. If we can reach the epistemic cognition, we will have a deep understanding of part problems.

Figure 2  Social Network

Social network analysis play a important role on researching the relationship, development process, prediction of information exchange in social network. We use graphs to present relationships between objects. Besides, degree, closeness, centrality are used to describe precise connection in social network.

Blog 2 for Social Media Analytics

After attending so many classes, I know more about analytical methods for content and sentiment. Jaccard index can be used for comparing the similarity and diversity between two documents. During this process, we need to consider term weighting. That's how we can determine whether a word is important. Maybe we find one word appear frequently, So it's more important. This approach is not completely correct, because there may be part of the words which belongs to structure words. Although they are used frequently in a specific document, we can't categorize this document by means of these unrepresentative words.

A typical processing technology shortly named TF-IDF is an information retrieval and text mining techniques for common weighting. The TF-IDF value increases proportionally to the number of times a word appears in the document, but is offset by the frequency of the word in the corpus, which helps to control for the fact that some words are generally more common than others. TF-IDF weight calculation method and cosine similarity often are used together in the vector space model, in order to judge the similarity between two documents.

Text classification make our daily life more efficient, we don't need to process all the text with the same procedures. Naive Bayes Text Classification is a good way for classification. In addition, K-Means Clustering offer an iterative refinement technique to measure distance of the similar points.

In term of sentiment analysis, actually we are in touch with many things and express our opinions everyday. How do we evaluate one object, one thing? Or how do we make a decision when we need to choose. In most cases, we don't pay attention to sentiment analysis. However, it plays an important role in collecting information to make decisions in fact. Furthermore, opinions are divided into various categories according to different norms, such as direct opinion, comparative opinion, explicit opinion, implicit opinion and so on. Sentiment classification methods include dictionary-based approach and supervised learning. The former is to check the polarity scores of the words in sentences to be classified. The latter is analogous to text classification.

This course requires some Python programming. Unfortunately, I have almost no programming foundation. Now I begin to do some Python learning. It seem a little hard for me to remember so many language rules. If someone who read this blog can offer some ideas about how learning coding well to me, perhaps I will learn Python more effectively. I look forward to mastering Python and do some application with it. Thank you for reading this blog. Please give some comments if you like. See you next time.