 Hello, welcome to this video session. In this video session, we'll be introducing the VECA with a learning outcome wherein after watching this video, we expect the audience will be able to explore VECA on their own. So primarily this video would be a hands-on session wherein I'll be actually be showing you the various interfaces of VECA and what all we can do through using VECA machine learning toolkit. So let us switch to VECA editor. So going back, let me switch to a VECA editor. So now I have a VECA editor open. Now the moment you open a VECA, you always see these five buttons to its right side here. Now let us quickly understand what is the use of these five buttons and what all we can do with each of these buttons. Generally, most of us, we always go with Explorer. In some little bit expertise user, they prefer either experimental or a workbench. And we do also have a simple CLI. So you do also see a command line interface here. So the advantage of command line interface is that it lets you handle VECA using command line interface commands which have been listed on the VECA documentation site. So mostly we are concerned with these four buttons, what are these four of these buttons? So when you go and open the VECA Explorer, what you see is in a screen where you can explore the various functionality of VECA. So that is a primitive one where you can load the data set from a file, ARFF file database and try out few clustering classification and some kind of machine learning and data mining algorithms. But the second button which is named an experimenter is more detailed one. What this button lets you do is if I click on this, it opens the screen wherein it lets you set up a specialized experiment with the help of various algorithms and data sets. So the object of this experiment is to compare multiple algorithms on a data set or compare multiple data sets with respect to one algorithm where you will get a very detailed result of comparisons. You can compare two algorithms where it's almost like an Apple to Apple comparisons. So it lets an ML developer to compare two algorithms for a specific ML problem and also at the same time let him save the results for future analysis and further conclusions. So let me quickly open explore and show you how the screen looks. This is how the screen looks when I open VECA Explorer here and when you click on open, this is how you load a data set and you can also load a data set from a URL and further our self-explanatory the generate lets you generate data for us. So generally by looking onto the screen under the pre-process tab, one of the very important to note down some very important is this ability to apply a filter on a specific data set. So if I load some sample data set, let me load a sample data set from the VECA installation itself. So I have my VECA installed on C drive, you will find a folder data under this. Let me do one thing. Let me load this iris data set. So when I load this iris data set, these are the attributes in that and this is a label. So it's a famous classification data set used for showing the demonstration of how the classification algorithm works. So that's how we load the data set and here this attribute, this window here, shows you some of the fundamental statistical properties associated with that. For example, I have selected a sample length and you can see that its minimum value means standard deviation, which gives a good amount of idea to the ML developer about the entropy present in this attribute. Which could be very much useful in deciding or which could be very much useful in feature engineering when the number of features are too many. So this is there, but one thing which is to be very much important is your filters. So here under filters, we have two types of filters, supervised and unsupervised. So what exactly is supervised and unsupervised filter? So supervised filters, they work on class labels. So whatever the filtering works here, they look at the class label and sometimes I want some filtering applied on one specific column or in a set of rows in your data set irrespective of what the class label to which that record belongs. So that time we prefer unsupervised filtering and under supervised and unsupervised you always find two more categories called attribute and instance base. Attribute related filters, they are some which work on column-wise. So they apply on an entire column and instance-wise are some which apply row-wise. So they apply horizontally on a specific set of rows based on what are the criteria which you give. So we will have a separate video session to demonstrate actually working of attribute base and instance base pre-filters. So when you explore unsupervised, here too you will find, in fact you will find more number of pre-filters on unsupervised because most of normal, most of pre-processing can be done on column independent of class labels if there isn't any much of relation between that. So some of the very famous pre-filtering is you can discretize the column values. You can convert a nominal to binary, binary to nominal and at the same time some of the very famous are normalizing. So what they do is they let you normalize entire column in a specific range. So there are many pre-filters available. So when you click on instance based there are few for example randomizing records or setting a subset or removing a duplicates. So these are one which apply at row level. So these are the pre-filters on that. So once I load the data here, what you next wise we move is we either try to classify the data set here by choosing few classifier algorithms or cluster it, association, select attributes are one which lets you do a feature engineering specifically like information gain or want to you want to know what's the gain ratio and here they let you visualize the data with the various helps of plot matrices. So this is general overview of explore screen. So once I close the explore screen and if I open experimenter window what you will see is you will see a completely different set wherein here what you see is here you set up an experiment and when you set up an experiment you add few data sets under this and then you add a few algorithms on that. So it's like the way like you can test one single algorithm on multiple data sets or one data set and you can test multiple algorithms. So the comparative analysis of this walking of algorithms either data set wise or algorithm wise it's generated and later we load it and we do a comparison on that. So this is used when we shortlist few algorithms to be applied to ML problem and then on we set up a detailed experiment to find out which algorithm is better and which is poor one and how internally they have an effect on that. So this is an experimenter we will have for the series of video sessions on each of these screens with the help of a good use case example. So after experimenter we have knowledge flow the knowledge flow is nothing but whatever you do under explorer and experimenter this lets you do with the help of GUI. So it's like a drag and drop interface where you can apply few filters. So you see what all whatever we could see there everything is available here and what you need to do is you need to just make sure this is a graphical way of designing that. So for few who are not much used to the way they we handle explorer and experimenter some might even try out their handset GUI based as well. So after that then we have a workbench workbench is far more powerful and it has a detailed setup of working with this and it's slightly better than explorer and it's far more sophisticated with more detailed result analysis and ability to have a better workflow based machine learning processes under that. So this is a workbench and when you click on simple CLI it lets you type WECA class names and let you invoke algorithms, configure and train data sets using completely command line commands. So this is WECA here so let me quickly switch back to our presentation and I have a small reflection quiz for you. So what do you think does WECA support normalization with the help of pre filters? What do you think is it true or false? You can pause the video and you can go back to earlier slides and try to guess the answer for this. But the answer for this question is yes it supports normalization. So normalization is nothing but it we bring a values in a specific range and sometimes it's a very important transformation on data when we are particular with specific algorithms which are very sensitive to the data range values. So as a further resource for you to explore I would advise or suggest you to go through this documentation of WECA wherein most of this information and the some of the instructions to repeat some experiments have been very well given there. So that's it for this video thank you.