Since big data becomes more and more important in our life. As a fresh graduate in Economics and Statistics, I’m eager to learn more knowledge about big data. Thus, I decide to participate the course Big Data specilization, created by University of California, San Diego, taught by Ilkay Altintas (Chief Data Science Officer), Amarnath Gupta (Director, Advanced Query Processing Lab) and Mai Nguyen (Lead for Data Analytics), they all work in San Diego Supercomputer Center(SDSC).
This specilization contains 6 courses as follows:
- Introduction to Big Data
- Big Data Modeling and Management Systems
- Big Data Integration and Processing
- Machine Learning With Big Data
- Graph Analytics for Big Data
- Big Data - Capstone Project
In this blog, I’ll share what I learnt about the first two courses, Introduction to Big Data, during the first week.
Before learning Big Data technique, let’s talk about the sources of Big Data.
Where does Big Data come from?
Big Data mainly comes from three sources: machine, people and organization. In the following, I’ll talk about them one by one.
Machine data is the largest source of big data, which presents the notion Internet of Thing(IoT). Think of a world of smart devices at home, in your car, in the office, city, remote rural areas, the sky, even the ocean, all connected and all generating data.
People are generating massive amounts of data everyday through their activites on various social media networking sites like Facebook, Twitter and LinkedIn, online photo sharing sites like Instagram.
Most of these data are text-heavy and unstructured, which bring challenges of working. This is certainly the case for big data and these challenges have created a tech industry of its own. Many big data tools are designed from scratch to manage unstructured information and analyze it, like Hadoop, Spark etc. I’ll talk about them later.
The last source of big data we will discuss is organization. How do organizations producd data? Each organization has distinct operation practices and business models, which result in a variety of data generation platforms.
Many organizations have traditionally captured data at the department level, without proper infrastructure and policy to share and integrate this data. This has hindered the growth of scalable pattern recognition to the benefits of the entire organization. Because no one system has access to all data that the organization owns.
Organizations are realizing the detrimental outcomes of this rigid structure, and changing policies and infrastructure to enable integrated processing of all data to the entire organization’s benefit.
While, how are organizations benefiting from big data? Let’s take an example of Walmart. They collect data on Twitter tweets, local events, local weather, in-store purchases, online clicks and many other sales, customer and product related data. They use this data to find patterns such as which products are frequently purchased together, and what is the best new product to introduce in their stores, to predict demand at the particular location, and to customize customer recommendations. Overall, by leveraging big data and analytics, Walmart has maintained its position as a top retailer.
As a summary, organizations are gaining significant benefit from integrating big data practices into their culture and breaking their silos. Some major benefits to organizations are operational efficiency, improved marketing outcomes, higher profits, and improved customer satisfaction.
Characteristics of Big Data (V’s)
Big data is commonly characterized using a number of V’s.
This refers to the vast amounts of data that is generated every second/minute/ hour/day in our digitized world.
In general, in business the goal is to turn this much data into some form of business advantage. The question is how do we utilize larger volumes of data to improve our end product’s quality? Despite a number of challenges related to it.
The most obvious challenge is storage. As the size of the data increases so does the amount of storage space required to store that data efficiently. However, we also need to be able to retrieve that large amount of data fast enough, and move it to processing units in a timely fashion to get results when we need them. This means their performance will drop.
As a summary, the challenges with working with volumes of big data include cost, scalability, and performance related to their storage, access, and processing.
This refers to the speed at which data is being generated and the pace at which data moves from one point to the next. This brings additional challenges such as networking, bandwidth, cost of storing data. In-house versus cloud storage and things like that.
Additional challenges arise during processing of such large data. Most existing analytical methods won’t scale to such sums of data in terms of memory, processing, or IO needs.
This refers to the ever-increasing different forms that data can come in, e.g. text, images, voice, geospatial.
The heterogeneity of data can be characterized along several dimensions. We mentioned four such axes here.
Structural variety refers to the difference in the representation of the data, like formats and models. For example, an EKG signal is very different from a newspaper article.
Media variety refers to the medium in which the data gets delivered. The audio of a speech verses the transcript of the speech may represent the same information in two different media.
Semantic variety refers to the method of interpretation and operation on data. We often use different units for quantities we measure. Sometimes we also use qualitative versus quantitative measures. For example, age can be a number or we represent it by terms like infant, juvenile, or adult. Another kind of semantic variety comes from different assumptions of conditions on the data. For example, if we conduct two income surveys on two different groups of people, we may not be able to compare or combine them without knowing more about the populations themselves.
The variation and availability takes many forms. For one, data can be available real time, like sensor data, or it can be stored, like patient records. Similarly data can be accessible continuously, for example from a traffic cam. Versus intermittently, for example, only when the satellite is over the region of interest. This makes a difference between what operations one can do with data, especially if the volume of the data is large.
Thus, data variety has many impacts like be harder to ingest, be difficult to create common storage, be difficult to compare and match data across variety, be difficult to integrate and management and policy challenges as well.
This refers to the quality of the data, which can vary greatly. Because big data can be noisy and uncertain. It can be full of biases, abnormalities and it can be imprecise. Data is of no value if it’s not accurate, the results of big data analysis are only as good as the data being analyzed. So we can say although big data provides many opportunities to make data enabled decisions, the evidence provided by data is only valuable if the data is of a satisfactory quality.
There are many different ways to define data quality. In the context of big data, quality can be defined as a function of a couple of different variables. Accuracy of the data, the trustworthiness or reliability of the data source. And how the data was generated are all important factors that affect the quality of data. Additionally how meaningful the data is with respect to the program that analyzes it, is an important factor, and makes context a part of the quality.
This creates challenges on keeping track of data quality. What has been collected, where it came from, and how it was analyzed prior to its use.
This refers to how big data can bond with each other, forming connections between otherwise disparate datasets. For a data collection valence measures the ratio of actually connected data items to the possible number of connections that could occur within the collection. The most important aspect of valence is that the data connectivity increases over time.
Thus, valence brings some challenges. A high valence data set is denser. This makes many regular, analytic critiques very inefficient. More complex analytical methods must be adopted to account for the increasing density. More interesting challenges arise due to the dynamic behavior of the data. Now there is a need to model and predict how valence of a connected data set may change with time and volume. The dynamic behavior also leads to the problem of event detection, such as bursts in the local cohesion in parts of the data. And emergent behavior in the whole data set, such as increased polarization in a community.
Voilà, here are what I want to share with you. In the review of week 3, I will talk about the process of data analysis and Hadoop.