Data Mining in Social Media Marketing

Advertising is an effective way to promote a business, product, or service. It is a process that can be used in multiple formats, with online, television, and video being some of the more effective ways in our generation. More than $500 billion is spent on advertising each year; roughly $4 billion is spent on Facebook advertisements alone (Kwiatkowski). Due to our generation’s high usage of technology and media, kids’ exposure to advertisements is at a record high, adding yet another demographic to a company’s potential audience.

In fact, teenagers and 

When we click on a YouTube video and an advertisement pops up first, our cursor hovers over the “You may skip this ad in 5 seconds” button, and the moment we can skip it, we do. If advertisements pertained to our interests a little more, maybe we would pay more attention to them. But with that comes a very common issue: the advertisements start to get “creepy”. Many people have mentioned times when they were talking about a product of some kind to a friend, and soon after, saw an advertisement for that exact same product. It kind of makes one feel like they’re being watched -- even stalked -- and that’s something most people don’t like. At the same time, however, the amount of data in our generation can easily overwhelm users, and they may struggle to find content relevant to their interests (Zafarani, Abbasi, Liu 2014). This is where the idea of data mining comes in. Our Internet usage is monitored to ultimately try and make it a better experience for us.

Data Mining

Data mining plays a huge role in the digital world today, and most people don’t even realize it. Even if they are aware of it, they may not be aware of what it is , or how it is used by many companies. Before we dive into how data mining is used, we must first define what data mining is. The term ‘data mining’ is used in literature in multiple ways, making the usage of the term seem inconsistent (Christen, Alfano 2015). There are actually several different terms used to reference a similar process; two of those terms that are seen most commonly are ‘morality mining’ and ‘social media mining’, which have both been coined to the process of data mining in different ways. Here is how the three terms are similar, but different. In broad terms, ‘data mining’ is the practice of examining large databases to generate new information. To do this, companies begin by analyzing hidden patterns of data and categorize those patterns into useful information, which is collected and assembled, to be used for business decision making later on.

There is a process for extracting information known as Knowledge discovery in databases (KDD) ; this process takes raw data from provides statistically significant patterns found in the data as output (Zafarani, Abbasi, Liu 2014). Ideal KDD would be able to evaluate trends that may be overlooked in other cases and, “if any information is found that is relevant towards the goal, subclasses or relationships may be indicated, which can then be used to further clarify and expand a database’s value” (Norton 1999). Data mining is an effective way for businesses to get information about their customers and develop more effective strategies to benefit them. Many companies have integrated data mining solutions into their database products (Mikut, Reischl 2011).

Social media is defined as a “group of Internet-based applications that build on the ideological and technological foundations of the Internet, and that allow the creation and exchanges of user-generated content” (Gundecha, Liu 2012). Social media mining can be considered a subcategory of data mining, because it deals only with data from social media - which still happens to be a lot of data. Defined, it is the process of obtaining data from user-generated content on social media sites and mobile apps in order to form conclusions about users and use that information for different functions, the main one being advertising. It introduces basic concepts and algorithms suitable for investigating tons of social media data; discusses theories such as computer science, data mining, social network analysis, and statistics; and uses these tools and more to properly mine information (Zafarani, Abbasi, Liu 2014). It is data mining on a smaller level, in a sense. Tons and tons of data is generated through social media every single day. Research shows that this will only continue to increase as social media continues to grow in popularity and usage. This could ultimately make social media the most efficient way to get information about potential consumers, if it’s not already. There are some downsides to this procedure though. Social media data is vast, noisy, and unstructured. This can cause issues for the companies that are mining, making the process more time-consuming and even confusing (Gundecha, Liu 2012). Facebook and Twitter - two of the most popular social media platforms - have reported that they get approximately 149 million and 90 million visitors every month. It is hard to handle the immense amount of data that is retrieved from social media platforms, therefore social media mining is still emerging, and there are currently just as many problems as solutions (Zafarani, Abbasi, Liu 2014). It is definitely a more modern idea, although data mining goes back before social media, but there is still work to be done in this subject.

‘Morality mining’ is a term that will typically come up with the idea of data mining. Morality mining occurs when data mining aims to disclose information on the moral values of individuals and groups (Christen 2015). When you think about extracting information about people, more often than not, you think about the morality behind that. To get personal information about someone, you’d have to invade their privacy. By legal definition, ‘invasion of privacy’ is the intrusion into the personal life of another without a justifying reason or cause (US Legal). This could lead that person to file a lawsuit, and that’s something that no one wants, especially a business. A model of “moral intelligence” may give businesses access to information that is liable to being misused (Christen 2015), information that they don’t necessarily need to be able to access. Morality and ethics play a bigger role in the practice of data mining than most people originally think.

Process of Data Mining

So now that we understand the concept of data mining — as well as two other terms that go along with it — we can now start to get into how data mining actually works. Data mining is a component of what is called the KDD (Knowledge discovery in databases) process, and often involves applications of data-mining methods that are repeated until the process is complete (Fayyad, Piatetsky-Shapiro, Smyth 1996). A simple overview of the steps that make up this KDD process is as follows: “1) The selection of the target data based on an understanding of the application domain and the goals of the analysis; 2) The pre-processing of the data. This involves data cleaning, noise reduction, deciding on strategies for handling missing files, etc.; 3) The data transformation involving (ex.) dimensionality reduction or finding invariant representations of the data; 4) The pattern discovery (data mining, technically) including selection of the appropriate analysis tools, parameters, and models, as well as exploratory analysis; and 5) The evaluation , interpretation , and visualization of the result” (Christen 2015). Each one of these steps can be broken down and understood even further. First, a business needs to understand their customer and client objectives. They need to know what their audience wants, even when they do not particularly know themselves. Developing an understanding of the application domain and the relevant prior knowledge and identifying the goal from the consumer’s viewpoint (Fayyad, Piatetsky-Shapiro, Smyth 1996) is very important to this process. On top of that, businesses needs to be aware of the current data mining scenario. They need to take resources and constraints into consideration, along with other significant factors. After selecting their target data and considering all of the factors (pre-processing), businesses should then define their data mining goals and create a target set. Data is collected from multiple sources and organized into a set, on which discovery is to be performed (Fayyad, Piatetsky-Shapiro, Smyth 1996). This part of the process is tricky and complex; data from different sources are more unlikely to match easily. After collecting the data, the business must find the properties of it using the query, reporting, and visualization tools from the earlier steps. The data quality should be ascertained based on results, and missing data should be acquired. The data preparation part of the process takes up about 90% of the project’s time (Christen 2015). During this part, the data is selected, cleaned, transformed, formatted, anonymized, and constructed. Next, data mining experts select and apply various mining functions because you can use different mining functions for the same type of problem. The experts must assess each model; there is a frequent exchange with the domain experts from the data preparation. Then, they must find useful features to represent the data depending on goals (Fayyad, Piatetsky-Shapiro, Smyth 1996). After going through the pre-preparation, and once the data has been cleaned, the KDD goals must be matched to a data mining method. They then start exploratory analysis and hypothesis selection: that is, choosing the data mining algorithm(s) and selecting method(s) to be used to search for data patterns (Fayyad, Piatetsky-Shapiro, Smyth 1996). After this step, the actual data mining can begin. This step can be made a lot easier for the user if the steps before it are done correctly.

Finally, in the last step, data mining experts evaluate the model. If it doesn’t satisfy their expectations, they go back to the modeling phase and rebuild it, changing the parameters until they reach the optimal values. Once the optimal values are reached, and they are satisfied with the model, the experts then decide how to use the data mining results, in ways such as exporting the results into database tables or into other applications like spreadsheets. The KDD process can be difficult to fully understand and time-consuming; sometimes two of the steps could loop a couple times before moving onto the next step (Fayyad, Piatetsky-Shapiro, Smyth 1996). Although every single step in the KDD process is crucial, the data mining step is the one that is most important.

Advantages and Disadvantages

There are both advantages and disadvantages to data mining. The use of data mining has significantly increased since the 1990s, thanks to the digitization of many processes in the business world, but also in our day-to-day lives (Christen 2015). However, there is a strong negative relationship between social media [data] mining and public life, and before we can turn that negative into positive, we must democratize data power in three ways: 1) To address concerns about the potential negative effects of data mining on the public, it must first be subject to greater public supervision and regulation; 2) To address the danger of new, data-driven digital divides emerging, the technologies of data mining must be available and accessible to the public so they can be used in varied ways; and 3) Given the contribution that data mining increasingly makes to how publics and public issues are represented, data mining could be used in ways that enable members of the public to understand each other, reflect on matters of shared concern, and decide how to act collectively as publics, therefore allowing publics to constitute themselves as more reflexive and active agents (Kennedy, Moss 2015). Basically, if the public had more access to data mining — rather than just businesses — the negative aspect may begin to fade. When people don’t know about something, or that something is kept hidden from them, they tend to view it as more of a negative something automatically, because they know so little about it.

Kennedy and Moss (2015) suggest that giving the public this kind of access could help people find connections with the people around them, and that could potentially give businesses an easier way to reach out to audiences in similar target groups. Another primary advantage that we see in relation to data mining is the idea of making advertisements and Internet experience more relevant to the user. On the side of dis advantages, one that is thought about often is that of criminals, and criminal abuse based on novel insight and information gained through data mining. It’s one thing for businesses to get ahold of private information to use to better target their intended audiences. But what happens when a cyber criminal gets ahold of that same information? The most skilled cyber criminals can use this information for a lot more bad — even harm — than good (Christen 2015). The other, more obvious disadvantage would be giving businesses access to information that they don’t necessarily need, such as medical files or financial problems.

Ethics and Morality

Both of these terms -- “ethics” and “morality” are frequent in articles that you read about data mining. On the ethical side, the actual term ‘ethics’ usually refers to the value of ‘privacy’ — a cluster concept that unifies a number of moral considerations in support of data protection (Christen 2015). There is a fine line between what should and shouldn’t be accessed, and most of this information is accessed anyway. There is also an issue with forms of representation. While the public needs this representation to be varied, this doesn’t always result in equal representation (Kennedy, Moss 2015). Data mining may also make information accessible that has the potential to change a person, especially if it gets into the wrong hands.

There are also some ethical issues when it comes to the advertising side of the spectrum as well. Advertising sets out to fulfill four major demands: Attention, interest, desire, and action (Balanescu 2006). A lot of advertisements can achieve this by including humor, novelty, or some sort of a guarantee to grab the attention of users. But in recent history, it has been proven that adding a scandalous element is preferred because it catches the attention -- and keeps the interest -- a lot quicker (Balanescu 2006). At what point, however, does this cross the ‘ethical line’? 

Women are oftentimes seen in advertisements as a way to grab the attention of users; in fact, 80% of advertisements use a feminine figure to promote whatever it is a business is advertising (Balanescu 2006). There are many other topics that arise when talking about ethics in advertising, such as: whether or not a product is harmful (tobacco or alcohol), community standards (gender norms or sexuality), and even if it’s ethical to advertise to children at all (O’Barr 2007). At the end of the day, ethics, when in relation to advertising, is quite difficult to understand. What some people think is unethical, others may be okay with. This poses a difficult decision for advertisement designers every single day.

When we talk about the morality side of data mining, a term that comes up is ‘autonomy’. This is the human capacity to shape our own moral biographies, to present ourselves in a way that fits our self-understanding, to reflect our moral careers, and to evaluate and identify with our moral choices without the critical gaze and interference of others, and without pressure to conform to the ‘norm’ (Christen 2015). It is often cited as “the ground of treating all individuals equally from a moral point of view” (J Christman). Regarding this definition, however, it is difficult to argue that all autonomous beings have equal moral status or that their interests deserve the same weight in considering decisions that affect them. This is due to the idea that the abilities required for autonomy vary across species too greatly (J Christman). Morality is very complex, but it is also very important. It involves norms, values, and virtues that have a natural history. These norms, values, and virtues help guide people with respect to what is right and wrong. Moral identity is one of the main focuses of moral psychology research. It connects moral cognition with constructs and theories of personality science, and helps us better understand moral behavior (Christen 2015). 

Did you like this example?

627

26