Which engagement level can we predict better on Instagram? A classification approach

In social media, likes and comments are key metrics for success.  If you can predict how many likes and comments a post will get, you can better manage your content strategy and create content that resonates with your audience.  There are a few ways to build a prediction model for the number of likes.  In…… Continue reading Which engagement level can we predict better on Instagram? A classification approach

Demystifying Language Models: The Case of BERT’s Usage in Solving Classification Problems

Newcomers to the field of Artificial Intelligence (AI) often see the term ‘language model’ tossed around when discussing Natural Language Processing (NLP) tasks without any proper clarification of its importance and usage in solving real-world problems. So, this tutorial blogpost aims at demystifying language models by defining what a language model is, describing the common…… Continue reading Demystifying Language Models: The Case of BERT’s Usage in Solving Classification Problems

Core Elements of a Web Analytics Process

Here is a description of a web analytics process with just the core stages. The core of a Web analytics process consists of four essential stages, shown below. Collecting data: At this stage, you collect the basic, elementary data. Typically, these data are counts of things or events. Processing of data into information: At this…… Continue reading Core Elements of a Web Analytics Process

Log Odds Ratio: Going Beyond Simple Term Frequencies to Characterize Textual Categories

Gaining insights from text-based data can be a daunting task, even when the data is labeled with ground truth categories and ready for usage in machine learning tasks.Researchers often rely on simple methods like the frequency of words in each category to understand the collection’s characteristics. However, this approach is not always insightful, as term…… Continue reading Log Odds Ratio: Going Beyond Simple Term Frequencies to Characterize Textual Categories

VLCP: A four-level framework for social media user engagement metrics

When evaluating the performance of social media content, user engagement is a key performance indicator (KPI). Specifically, the interaction by the user with the content, such as viewing, liking, commenting, and sharing. From the mindset of a social media content producer, users with more public expressiveness of their interest have higher impactful engagement with the…… Continue reading VLCP: A four-level framework for social media user engagement metrics

Big Data Fallacy

Big Data Fallacy. The law of large numbers argues that the sample’s mean approaches the sample population’s actual average as a sample size increases. This concept is often, either implicitly or explicitly, taken as a justification as to why ‘big data’ (i.e., millions or billions of samples) cannot be wrong. However, there are contrary arguments and…… Continue reading Big Data Fallacy

Flaw of Averages

When working with numbers representing many people, one often employs the average to describe the people represented by the numbers. However, this approach can lead to serious problems due to the flaw of averages, which is that findings based on the average are wrong on average. In fact, often when dealing with people, the average person…… Continue reading Flaw of Averages

The Illusion of Data Validity: Why Numbers About People Are Likely Wrong

This reflection article addresses a difficulty faced by scholars and practitioners working with numbers about people, which is that those who study people want numerical data about these people. Unfortunately, time and time again, this numerical data about people is wrong. Addressing the potential causes of this wrongness, we present examples of analyzing people numbers,…… Continue reading The Illusion of Data Validity: Why Numbers About People Are Likely Wrong

When mere correlations are not enough: The Granger Causality test

In most data science-related problems, datasets consist of multiple variables, in which independent variables might depend on other independent variables. When the variables in datasets represent observations at different times, we call this dataset a time series set. The time interval in these data sets may be hourly, daily, weekly, monthly, quarterly, annually, etc. One…… Continue reading When mere correlations are not enough: The Granger Causality test

SIG-KM International Research Symposium 2022 Keynote

Super time at the SIG-KM International Research Symposium 2022, hosted online by the University of North Texas. Thanks to Jeff Allen and to Lu An for the invitation! Thanks also to the Center for Studies of Information Resources (CSIR) of Wuhan University (WHU) and other academic sponsors for their support of the symposium!