 In this LED we will talk about Pre-D2 learning analytics although we mentioned that in week 4 only we will see Pre-D2 learning analytics, we realized that in week 4 there are too many activities and we will be talking about two different models. So we thought of moving the last LED to about Pre-D2 learning analytics in week 3. What is Pre-D2 analytics? We briefly touched about this in a week 1. Pre-D2 analytics that is to analyze the current and historical data to predict the future events. So in this Pre-D2 analytics we will involve machine learning and data mining tools in the learning analytics. Pre-D2 analytics is very common, it is applied in different domains for example finance, health sector, telecom and other domains. For example the weather prediction is by done by Pre-D2 analytics, it collects a lot of historical data over the period of last year, last 10 years then it predicts what will happen next. Whether it will be sunny or rainy or next day. Also for fraud deduction in bank transactions can be done using Pre-D2 analytics. Another example is in a day to day life is having anti-spam machine email. In an email we collect historical data about what of the words the spam email contains then we create a model then we will predict whether the next email is spam or not. Although it is not predicting the future event but it predicts the incoming email before the learner or the user detects it is a spam. Pre-D2 learning analytics, so what is Pre-D2 analytics and for learning? In Pre-D2 learning analytics it is to understand what might happen next in the students interaction to the students. Whether the student is able to answer the question correct or student needs more help in the next slide or the student will need, will pass the exam or student needs hint in the next question those things will be predicted. In order to do the Pre-D2 analytics or to predict the future event of the students or learners interaction we need to record the historical behavior of the student. From the historical data we can create a model then extrapolate that model to predict the future events. Pre-D2 analytics involves descriptive and diagnostic analytics also. For example, first we need to describe the data, what are the data then we need to identify the pattern that is the descriptive analytics or diagnostic analytics then we need to extend or extrapolate that model to predict the future events. Let us see the activity. In this activity we discussed that the Pre-D2 analytics needs historical data to predict the future events. However, do we need historical data of a same learner or we can use data from other learner. For example, as I mentioned in weather prediction we realize that in order to predict the temperature tomorrow or in order to predict the climate tomorrow we need to understand what was the climate in last one month or one year or last 10 years in different seasons but it is about the same climate and same area right. So, in learning analytics in order to predict the future events do we need a data from the same student or can we use the data from other learners. If you say we can use the data from other learners what are the restrictions, what are the assumptions please think about this question write down your answers after writing down the answer you can resume this video. You might have realized that we may not able to collect enough data of a student to not to predict the future events because a student will be newly starting the learning environment in order to give a recommendation or prediction based on that you might need to wait for a one week or one month so it is not possible right. We need to give a personalization or adaptive feedback hint messages immediately after they join the learning environment program. In order to do that we need to consider the data from other students or from the similar learning environment. So, if I say similar data what are the similar data, what are the restrictions in order to collect the similar data for example it may be domain. Is the student the data we collected is in the same domain is in a math or the same concept we need to think about it also if the same learning environment we cannot apply the data collected from learning environment to x on learning environment to y. So, we need to collect data from the same learning environment also the interaction behavior should be similar for example in learning environment to x first version you might have different menus or different interaction behavior allowed for students in second environment you might have a additional features that is not allowed. You need to collect data which has a similar domain similar interaction behavior also similar learning environment. If you are talking about technology enhanced learning environment for example if it is a MOOC or a classroom environment we need to provide a personalized learning content based on the similar user data taken this course in last semester in a MOOC the student will be taking the course for first time but we have a same set of students or similar kind of students would have taken the course in MOOC. All the data we collected from the MOOC platform will be applied used to predict the future events of the current students. So, we can collect data from the students who have taken this course in the previous semester or previous year or previous events and create a model and use that to predict the future events of the current students. So, who are the stakeholders and why were we doing this predictive learning analytics? What is this what is in for stakeholders for predictive learning analytics? For learner it is very important because based on this predictive learning analytics model we are going to provide the suggestions of feedback to improve the learning. So, it is very important for learner as a stakeholder. The second stakeholder we know as teacher for teacher it is very important to know what will happen to student next whether student will able to solve that particular problem or will be passed the exam or the student will be able to complete the given task. In order to help if they are not able to complete task the teacher has to know when to intervene and what to intervene. So, the predictive analytics should give a insight saying that the particular student in your class might have a problem because this learning activities or interactions with the systems or environment is not enough it says that the student might not able to complete the problem given to them. For example a teacher can know that in the class out of 10 students 3 or 4 students may not submit assignment in next month. So, what can we do? So, teacher can think of other different strategies and make motivation to the students and ask them to submit assignments on time or they can extend their assignment deadly. So, similar activities are helpful for teacher. So, predictive learning analytics is very important for teachers as a stakeholders to understand which student will complete the course also to understand what is the problem in the current learning content. We talked about other stakeholders that is institutes, academics, academic analytics stakeholders or the institutes or districtates. They also will be interested in about predictive learning analytics for example what will happen which course will be taken by most students in next semester which course will not be taken by most students in this college. So, they can decide whether to run a course next year or not. The other important stakeholder in learning analytics is content developers of the eLearning systems. If the learning analytics or predictive analytics says that the students will have a problem if their interaction behavior is kind of certain pattern then you might need to give this particular recommendations. These informations will help the students to create the automatic agents. In eLearning systems, we can have agents which can interact with students based on the learner behavior. In order to design this agent, monitoring the students behavior from the historical data is very important.