Recommendation Systems used in Popular Website

We learned the basic knowledge of recommendation systems last week. Living in an era of information explosion, we have so many options that make it even harder to choose which to bye,what to do or whom to know. Due to the factor that we usually rely on some suggestions and recommendations when we make decisions, here comes the recommendation systems to solve the problem.

I'm already interested in recommendation systems when I started to use the website like Taobao, Weibo,Facebook,LinkedIn,Amazon...These popular websites adopt different method to realize recommendations.

Amazon's Recommendations

Amazon uses recommendation systems to recommend new products to users.It wants to show us with products that we will be interested in seeing.Amazon's method of recommendation is kind of complex,but it is centered around the products a user views and purchases.They utilize a system known as item-to-item collaboration filtering.

Amazon's Item-to-Item Collaborative Filtering focuses on the items rather than the users. It does not try to match the user to similar customers. Instead, it matches each of the user's purchased and rated items to similar items, and then it merges the similar items among them into one recommendation list unique to the user.

YouTube's Recommendations

Youtube utilizes recommendation systems to bring videos to us that we will be interested in. They are designed to increase the numbers of videos we will watch, increase the length of time we spend on the site, and maximize the enjoyment of our Youtube experience.When I watch the video on YouTube, I can't stop opening the next one in the recommendation list.

In order to obtain personalized recommendations, Youtube's recommendation system combines the related videos association rules with the user's personal activity on the site. This includes several factors. There are the videos that were watched - along with a certain threshold, say by a certain date. After all, you don't want to count videos watched from 2 years ago if the user has watched enough videos, most likely. Also, Youtube factors in with emphasis any videos that were explicitly "liked", added to favorites, given a rating, added to a playlist. The union of these videos is known as the seed set.

LinkedIn's Recommendations

LinkedIn is a business-oriented social networking site, forms recommendations for people you might know, jobs you might like, groups you might want to follow, or companies you might be interested in. LinkedIn uses Apache Hadoop to build its specialized collaborative-filtering capabilities.

Basic knowledge involved in Social Media Analysis——Machine Learning

We were all impressed by the smiling face Rosanna drew in the class.It showed us how to describe happy and unhappy using mathematical language as well as the knowledge about machine learning.

Machine learning is a subfield of computer science and statistics that deals with the construction and study of systems that can learn from data, rather than follow only explicitly programmed instructions.

It's all about the question "Can machines do what we (as thinking entities) can do?" In fact, learning algorithm is so powerful that we are unaware of using it everyday. If you use a website like Taobao or Amazon that will often recommend clothes and book to you,these are examples of learning algorithms that have learned what sort of things you like to buy,and can therefore  give customized recommendation to you.
Machine learning tasks can be of several forms:Supervised Learning,Unsupervised Learning,Reinforcement Learning.

Supervised Learning

The theory mentioned in the class is supervised learning. If we have some documents which we know their sentiment class,we can use them to train a classifier.

In supervised learning, the computer is presented with example inputs and their desired outputs, given by a "teacher", and the goal is to learn a general rule that map inputs to outputs. 
Regression problem:
The variable you are trying to predict is a continuous value.

Classification Problem:
The variable you are trying to predict is discrete rather then continuous.It's either 0 or 1,just the same as happy or unhappy.In fact ,more generally,there will be more learning problems where we'll have more than one input variables and features.And one of the most interesting thing is what if you data doesn't lie in a two-dimensional or three-dimensional or even a finite dimensional space.But it's possible what if your data lies in finite dimensional space.It turns out there is a successful classes of machine learning algorithm called support vector machines which actually takes data and maps data to a finite dimensional space.

Learning Theory

It's about how and why learning algorithm works.We try to understand what algorithms can approximate different functions well and also try to understand things like how much training data do you need.If you're trying to design a learning algorithm,should you be spending more time collecting more data or it is the case that you already have enough data so don't have to waste time.

Unsupervised Learning

In an unsupervised learning problem,you're given a dataset without right answer on any of your data.Would you find any interesting structure in this dataset?The clustering algorithm is one example of unsupervised learning.It turns out these sort of unsupervised learning algorithms are also used in many problems--things like organization computing clusters,social network analysis,market segmentation,even for astronomical data analysis.

There is a classical problem called cocktail party problem which use ICA algorithm to separate out interested voice from all this loud background noise.

Reinforcement Learning

This refers to problems where you don't do on-shot decision-making.In reinforcement learning problem,you are usually asked to make a sequence of decisions over time.

 

Personal View on Social Shopping

Some of you may feel tired as the epidemic of commercial shopping broke out through the social media especially in WeChat, yet you can’t ignore the power of social shopping for the Hegel's "what exists is reasonable"

Here I list some potential reasons for the popularity of social shopping,

1. Cheap cost

For the majority of the people who care about their budget for opening a real shop, social shopping sounds like a good choice if they have a large circle of friends. They can broadcast their commodities through the convenience of social network without any cost.

2. The technology support of convenient payment

Such as the trend about Alibaba, the success of electric commercial due to the convenient payment to some extent. And make the payment easier for the customs is important for it’s fashion.

3. Amount supporters

The social shopping can prevail quickly among the friends, so it can be generalized because the relationship between the owner. Forwarding is a great method to promote the merchandises.

4. Free style of trade

It can be realized that everyone can be merchant and customer at the same time. For someone have something to sell he can post them on the social network, and he can get something he likes on the social network at the same time.

As listed above are personal opinions about the social shopping. For reference only.