 Welcome back to learning analytics course. So, I mentioned in couple of videos ago that LA can be also defined as applying analytics on learning data. If we take that definition, this consider that analytics is another circle and if you are applying analytics on learning data that can be called learning analytics. Let us focus on that definition and start applying that definition in our course. Since we can apply the types of data analytics in learning analytics also. So, we organized our course in considering the types of learning analytics then in each week we describe about each type and we will work with the demo and data collection in each types of data analytics. What are the types of data analytics? There are four types descriptive, diagnostic, predictive and prescriptive. You might be known these types when if you have done some data analytics or static scores in a different names, but these are the basic types in a data analytics. What is descriptive? Descriptive is like what happened, so from the data what happened will be described by tables or charts or written words or some graphs or something like that. From those data asking why this happened is diagnostic. And if you know why it happened, asking what will happen next is a predictive. And if you know what will happen next can we change the learner outcome, can we help the learner to achieve his goal, can we make it happen like what we want it to happen like what we want learn to learn, learner to learn. So, let us briefly describe this four terms and in this course every week we will talk about one type of analytics. Imagine that you are going to a doctor and the doctor ask you first questions what happened to you. So, you described your symptoms. So, describe is actually that you describing what happened to you. And based on that the doctor is asking other questions why it happened? Did you eat anything last night or did you have a milk with in the coffee whether you have something fried food. So, based on those informations like comparing your symptoms with other informations you might have done the doctor is trying to come up with the model and trying to say why would have it happened. So, it might have happened because you might have a stomach infection or you might have a fever. So, doctors already have a model for each symptoms and other indicators there are such a type of causes like if you have a stomach ache you are having some bad food or stale food you might have a stomach infection or you are having fever or something like that. So, doctors have a basic data of yourself and also can able to predict what would have happened to you that is a predictive model. So, based on that doctor will prescribe us tablets for you to get well so that your body conditions get better. It is a simple analogy of LA first you described your symptoms to doctor then doctor diagnosis then doctor predicts what would have happened most cases this prediction is correct in some cases a prediction may go wrong that is why in initial screening you may not get cured then they go for second test and multiple test and try to identify why the exact cause. So, based on that they prescribe some medicines for you to cure. So, you can take the simple analogy of you visiting a doctor or hospital scenario for the learning analytics look at how it can happen in a first video we had an example that you are a teacher and you are teaching the same course the last 5 years of 30 students or something like that and you have access to the students data such as academic background profile performance whatever data you can imagine all the data you can collect from the students you have it in a classroom environment. So, what data will you report and how do you report? I want you to pause this video and think about what data you will report. It is not the one class suppose we have 60 students in year one and you might have teaching for last 5 years so 3 and the students got data. So, suppose for example 2015, 2016, 2017 so each year there is a different set of data for 60 students. If you take performance there will be like 60 students into 5 years so 60 cross 5 that is something comes up. So, imagine this data if you are collecting more like a performance in the midterm or performance in each questions or the profile imagine this data. I want you to think this data and how do you like to represent? Will you represent a bar chart? Will you represent in pie chart and why? Please pause this video, write down your answers. Once you write it down, resume this video to continue. Consider you can have a lot of data like attendance, performance over the years, assignments course, the class feedback, their entry test or exit test all these information you might have it. So, these are the data you have over the last 5 years. Now you can might represent using pie chart. Pie chart is may not be common for all the data representation. Pie chart to represent what happened in a certain percentage compared to other set of data or histograms is a nice one. Online graph or scatter plots you can use different kind of plots to represent this data. Simply representing the data in the tables is not enough because this data is not showing you the tendency or this statistical information. Or you can use plots or you can use some other graphics for representing your data. If you represent this data, this can go in part of a developed dashboard like a dashboards where teacher can access to students data, students interaction with the systems everything can be there. And these dashboards are generally used in an academic analytics. For example, for very large data in a classroom environment of 60 we might not use dashboard. But if you have students say 1000 students interacting with say 30 videos and they have done a 3, 4 exams and lot of questions then dashboard is better. So, what is descriptive analytics? It is describe or summarize the raw data and make it something interpretable like you have to represent all the information of the data in a single picture. And it also provides the facts and the summary of the data in a single slide or single image or single picture. And it is easy to understand by different stakeholders. This is how you have to describe the data. It will also help in finding the trends in the historical data. If you have fast 5 years data you might say the students in 2018 did not do well in their exams compared to other group. You might have reasons why it happened. And this describe descriptive analytics data, the graphical representation should answer what has happened. If someone asked what has happened in 2018 exams score you can just simply look at the graph and say yeah 68 percentage students passed more than 80 marks something like that. So, this scripted analytics should tell what has happened from the checking the tables or figures or some other visual representations. If you want to show the scores of students in different tests. So, you can use the descriptive analytics. For example, say there are 4 students and they have taken a test each like January, February, March and April. You want to show the students score over last 4 months and you want to compare amongst them. You can show a simple bar graph. Or you want to see the transition of students performance in 2 tests. For example, you conducted entry test and exit test. And you want to show the transition of the students performance over entry test to exit test. For example, the group of students who got this score and they might have got a better score or least score in this test. So, this kind of transition also can be represented using stratified diagrams. This diagram is created using a tool called ISAC and we will talk about this tool in a third or fourth year of our lecture. So, let us move on to the next topic, next type of analytics. Now you have collected the data from your past course and also you decided how to represent them as a bar chart or line graph and you know what data you want to represent. As a course instructor, what you would like to analyze from this data? Like what question you want to answer from this data? As I mentioned, suppose you see at 2018 here, performance of students in your class has decreased compared to other class, other years. What question do you ask? Why that particular years performance has been down? How do you answer that? Please write down your answer in a paper, pause this video. After writing it down, resume the video to continue. So, you can ask questions like why a student dropped out of a course? Or it is a one particular student or why a student failed in an exam? Like that the student might have failed in a particular exam, why the student failed in an exam? Or why a student has a high failure rate or why a school has high failure rate? For example, a school in a district A has a very high failure rate, the performance is very poor. Why that school has very high failure rate? So, diagnostic analytics is actually, you start looking at why it happened, why this particular school or particular year, particular student is not performing well or why the student is performing really great, why the school is achieving 100 percent success all the time. Then you might need to collect other variables associated with that and you might do correlation and regression on that. So, what is diagnostic analytics? Diagnostic analytics deals with the process of finding out the reason that lead to an event in the past by making use of descriptive analytics. From the descriptive data and trying to find out why it happened, why this particular case, why what is the reason begin this is a diagnostic analytics. Is it answers the question, why things have happened? Then it also says that why things have happened in this way compared to the other ways. Suppose you have teaching two subjects in last three years, in 2018 the same set of students would have done a better in one subject but bad in other subject. Why things have happened in one way compared to the other way in other classrooms? So, you can have a lot of questions coming from the why. So, what we do, we basically apply some techniques for the diagnostic analytics, for example, correlation. Suppose consider that in your class in 2018 year batch students performed really bad, you can correlate with the students attendance rate. If the students attendance rate is correlated with the performance, you know that the students has a poor attendance rate in that particular year or there are a lot of holidays in that year. So, the student they have not able to come to class and you are not able to teach cover the topic. So, their performance is gone. That is a one chance. Or if you have a multiple variables like the students assignment scores or the sports meet happen in that particular year or the attendance rate, the background information profiles, you have multiple informations then you can render regression analysis. Or you can do a pattern mining if you are using the environment like a technology enhanced learning environment instead of a classroom environment. If you know the students is interacting with the system and a fixed set of time like they are reading and they are doing some analysis, they are interacting with the simulator, they are answering some questions. If you know the students set of actions already in intelligent learning environments, then you can collect this data, actions data, then you can run a pattern mining or you can also apply the process mining on this data. This will answer the question of why something happened. That is called diagnostic analytics. So, in this video we saw couple of things. One is descriptive analytics and diagnostic analytics. We will describe each of this in detail and the tools we mentioned in this video in the coming weeks. Thank you.