 Hello everyone. I welcome you all to this introduction course to data science. We will start with the famous and renowned quote by John Owen. Data is what you need to do analytics and information is what you need to do business. So as said if you are doing your data analytics job data crunching data munching and you are using the data to derive the business and unless and until it is not generating the value it is of no use to business. So you can use the data to uplift the business to get your sales revenue increased by some amount to get more opportunities to your pipeline that's where the data science being contribution is and these days we are seeing that AI and data science is disrupting each and every industry whether it is banking BFSI banking finance and insurance data science is getting used whether it is healthcare whether it is aeronautical space whether it is IT and whatnot. So it is widely being adopted in each and every industrial sector be it manufacturing be it like non-code job or whatever everything is getting replaced by data science. So what do you expect from this course? Is it a definitions? We will look at what is data science? What is machine learning? What is deep learning? No this course is not about reiterating or the things that are on google one can easily find the definitions and reiterating those in the course this is not that course okay. So we will go on a journey a journey that is meant to understand what is data science by making you contemplate by making you imagine the scenarios what I will describe. I will make you go through that journey and like you will imagine all those scenarios in your mind and you will like get to know that this data science concept is nothing new it is very very old concept which is getting widely adopted these days and how that boom happened how that things are shifting towards the data driven decision and things like that. So I just need your strong imagination power we will set a journey and you will see a bigger picture and you will understand all the concepts what you after doing this introduction course to data science you will understand what is data science what is machine learning what is deep learning how they are different and what are the different kinds of data is available to us and also you will get to know about the roles differentiation because mostly people are getting confused or they are mixing up some roles which is predominantly in the data science domain such as data scientist data analyst business analyst business intelligence and whatnot. So we will see each and everything in this introductory course which will clear your old questions that you have related to this field of data science. So let's get started. I welcome you back to this course and we will start with some basic questions that you are also getting confused while studying or while reading about the data science what is data science who is data scientist what does he do what is the role of data scientist and things like that. So what is the difference between data scientist and data analyst and the business analyst and everything we will see at last and it will all get clear what questions you have in your mind right now and it will get solved. So starting or going to our next slide this is usually you will find in every introductory course which will start with the definitions what is AI what is machine learning deep learning reinforcement learning and the computer vision and they will just explain the meaning of that but this course is not like that I will not reiterate the definitions that one can read from the Google and can understand will not start with the definitions rather I want you to come like set a journey for you and imagine the things that I want to do right now. So I just want to do a simple task a very very simple task of yours just imagine your daily routine what is your daily routine looks like and it's it should be starting up with the morning routine okay so I'm taking you or setting up a bigger picture for you so just imagine what is your daily routine after you get up in the morning okay so we'll start with that so if you are a person who wakes up in the morning and you do either of these three things and I can bet like all three like all these three things that I have to find like 95 percent of the people will fall under these three okay so just imagine you wake up in the morning what is the first thing you you can either like speak speak in the sense yeah you can like wish your good morning or good evening or your parents can wish you the same and you being the listener so either you can speak to someone your roommate your friends your colleague to anyone that is the first thing you will like wake up in the morning you attend a call and you will speak so that is you will have what kind of data that is what more important that is the speech data okay like you are speaking to someone or you are either listening to someone second scenario you wake up in the morning you just like just had your phone in your hand and you read some of your notifications that you have you read you can read your emails you can read your tweets you can see your instagram posts and things like that but whatever okay so what kind of data is involved in that and you have guessed it right that is textual data you are reading something you are reading a mail you are reading a tweet you are reading your LinkedIn post everything that you see there is in form of textual data the third point third scenario you pick up your phone and you just click the picture or you just uploaded a picture on the instagram or you just open the instagram and scroll through the images so that is the third kind of data that you are seeing that is either a video or a picture and I can bet 95% people are using these three dominantly like there is nothing more classification of data a data can either be in speech format or you can say the audio format either it can be in textual format number format number is what we can calculate we can do operations and the textual is something which we have as a textual data that is sentences and third one is we can have either images or we can have videos so these three scenarios or these three categorization of data is only possible nothing else your data is divided into these three bigger chunks and everything that you like read everything jargons every other word you will read about the data will be categorized into any of these three hi I welcome you back to this course so now we have understood that the three major categorization of data we can have is either you can speak or listen that is your audio file involved or audio speech is involved second one is either you can read a text you can write a text you can write a number numerical data you can perform calculations so that is nothing but your numeric data or textual data the third categorization is you can either click your photos upload the photos watch the videos on youtube and things like that so that is your third categorization of data so you have understood that we are like from morning till night we are equipped with these kinds of data nothing beyond so we are we will speak to our colleague to our friends to our bosses we will read something we'll write something we'll watch something and we end our day so these are the three data points that are human generates at each and every second of his time correct now coming to the definitions part so that it could make you more clear the first one for the speeches speech data or you can say for the in what we call these as a smart assistance or we can call it as intelligent virtual assistant in the language of data science IVA in short form so the definition states that artificial intelligence is intelligence demonstrated by machines unlike the natural intelligence displayed by humans animals which involves consciousness and emotionality so these days we are having these devices called Alexa Siri Google now and Cortana which can understand the human speech you just ask the question to them and they will answer you in the natural language you ask them what is the weather today what is the score of the match and they understand each and every word that you speak and they comprehend that speech into their back end system into the text they will get an answer and can get back to you within speech so this is like evolving these days and you can do endless things using your Alexa you can control your home you can control any operations using voice you can listen to music and things like that second one the natural language processing this is another definition I can say the branch of data science which is natural language processing and as the name suggests you are processing what a natural language a textual form a written form which you have in the data so natural language processing is a subfield of linguistics computer science and artificial intelligence concerned with interaction between computers and human language in particular how a program computers to process and analyze large amounts of natural language data so these days this kind of data is getting widely adopted because in one study or you will see in my next slide in Gardner study it is mentioned that 80 percent of the data is mostly in your unstructured form and what I mean by unstructured is it is either in terms of or in form of textual or in terms of video or photos or things like that so that all are your unstructured data and 80 percent of the world data okay everywhere data is natural language so what is an application of natural language so you can derive a sentiment from the comments you must have bought something from amazon and flipkart and you see people reviewing that item so you can take a natural language response over there you can build a model and you can have a sentiment whether the user is happy or not and companies are really using those sentiment analysis in everywhere because you can find a review in each and every website either you book your hotels on make my trip either you book your houses on air bnb everywhere you find option of review and based on that review you have you we must take a decision okay we do that if a property is rated four star five star we usually read the reviews before booking anything we usually read the reviews before buying anything so that is the power of natural language and you can take it to next level you can build your quotient answer module you can take your like you can bring your sentiment analysis you can do your neural machine translation there are various languages you want to translate from one or the other you can do speech analysis and whatnot coming to third point of images data so data image processing is used of a digital computer to process digital images through an algorithm computer vision is interdisciplinary scientific field that deals how the computers can gain high level understanding of the digital images and videos so we have images we have videos which is nothing but frames of photos only so videos are nothing but the frames of your pictures only so you can process these images you can make a conclusion out of that you can even detect in a healthcare you can detect cancerous tissues or tumors which are not visible to a human eye but a machine learning algorithm can differentiate those patterns and can get whether the person has a cancer or not you can use images for self-driving cars for detecting any like anomaly in the product and things like that so all three of these data and like whatever you will see now it's the application of these three data either you are using speech data which is governed by Alexa Siri and these virtual assistant either you will use the textual data to do any some sort of solving some problem or either you will use the images data to again solve some bigger problem or bigger challenge so I hope this makes you understand what is happening in actual in data science world and everything that you see around is one of the other application of these three data hi I welcome you back to this course so now let's get started of the we have set the bigger picture now let's see what is data science and the data categorization so data science is an interdisciplinary field that uses the scientific methods processes and algorithms so this is nothing new algorithms are existing from I guess past 20 or you can say past 50 years when the you can say the when the world of software engineering hit a market so algorithms are existing from long back and it's just that we are using the application of algorithm to solve a problem of data science so and if you continue to definition it's a algorithms and systems to extract the knowledge and insights so that are like really two important words you what you want to do with the data what is the need to do data analysis or data science you need to extract the knowledge and insights so that is the whole purpose of doing data science you have to extract knowledge and insights from any kind of data that is it can be structural data or it can be unstructured data and data science is related to data mining machine learning and big data so we will see those dark ones in terms so now the term one use is you have to find the knowledge and in terms of data science we call it as KDD knowledge data discovery in terms of data mining so that is what we want to achieve and find the insights now coming to the data categorization we can either have the structured data or unstructured data now let's see the difference between those so the structured data is quite simpler one or you can say that is being used from like past several years you have heard of excel files you must have used your excel files so that is nothing but your structured data structured data has the rows and columns so you have those kind of tools you have excel you have sequel databases in which you have a table like rows and column structured data that is stored and you can do anything based on the data you have so if it is like sales data revenue data you can find average you can sum all those you can do any kind of perform like statistical analysis on that kind of data so that is something which we already know which we have worked on structured data on the other hand the unstructured data cannot be represented in the form of rows and columns and relational databases so we have special kind of data to store our images videos and which is like no sequel databases okay coming to the second point so as we have said it can be of number dates and strings we have like sales category what items are getting sold on each day so we have dates we have string as a product name we have numbers as a quantity that is being sold by as an example for companies data the on the other hand in unstructured we have images audio video world processing file emails emails is nothing but your again textual data that is coming spreadsheet etc and as I highlighted this fact in my previous video so estimated only 20% of enterprise data is in structured form and the rest 80% is in unstructured form and that is quite evident like mostly if we see around us all the knowledge is present in terms of textual format so textual data is predominant in terms of when we talk about any industrial data so we read a lot of information on the internet we write a lot of information on the internet so all is getting stored or all is getting processed in terms of textual data coming to the fourth point obviously structured data is an numeric and integer and strings so that requires very less storage so that is like less extensive to the compute resources on the other hand your unstructured data require more storage more space in your heart is and this is quite resource extensive the last point is easier to manage and protect the legacy solutions so we can have like data privacy we can have the unlocking of we can say we can save our files in such that it is not getting unauthorized access and things like that but the textual data is difficult to protect or we can say that is widely accessible to the people around you or around any social media platform so your textual information your photos your videos are quite prone to data privacy or you can say cyber threats hi and welcome you back to this course so we have seen and set a bigger picture of data science we have seen structured versus unstructured data so now coming to this very famous pyramid of data science is the step or you can say the procedure to reach at the wisdom level so simply data science is a study of data and it is often confused with the data mining so in terms of data mining or data extraction that you want to do a performance analysis the first is kind of a raw data which is of no use to business data is being collected at various stages of the company or you can say various stages of the industry it can be collected on sales level marketing level on product level what is the issues behind the product you can have the logs data you can have the web analytics social media analytics and whatnot so you have the raw data in terms of anything okay like that is widely being captured by the industries now the second step after capturing or after having the raw data is to find the information out of that data information is nothing but you are using the relevant data and discarding the rest so you are using what is what is relevant for your business use case that you are solving and the rest you are discarding and after the just figuring out which all data or which all you can in terms of rows and column you can imagine which all features or column you have in your data you call it as your information and then using that information you are deriving some knowledge okay so deriving some insights some knowledge that is again helping your business to uplift and increase some sales or revenue or do some better performance so when that is achieved that is you are going to the stage of wisdom and you are achieving that criteria so let's come to our next slide so now we have a discussion on machine learning a famous word these days that is being used widely across industries so machine learning is a study of computer algorithm that improve automatically through experience it is seen as a part of artificial intelligence when we say that the algorithms that improve automatically through experience so I will give you an example that you can correlate with the machine learning thing imagine a newborn baby okay so he knows nothing about the world it is just he's a new newborn and he's just came to this world he will or she will eat whatever is feed it to him he will pick up anything like whatever you give to a child or a baby he will just pick up he doesn't know whether this thing will hurt him or whether if he touches the fire will hurt him or not he will do out of the curiosity he will do but once he put his finger on the fire he'll get some pain that will come back to his mind as a pain that is not a rewarding thing and that is the experience he is getting stored so a child when learns anything new he must be learning how to walk how to ride a cycle he like many times he will fall down and he will learn how to ride like learning through experience so that is what machine learning is a machine learning algorithm learns from an experience learn from the data that is fed into the model so this is the concept of machine learning and through this concept we have divided the machine learning into three major categories supervised learning unsupervised learning and the reinforcement learning so we have seen three categorization of machine learning algorithm now we will discuss each one of them and we'll see what's the difference among all three so first we will start with the supervised learning so as we have said machine learning algorithm learns from the experience learns from the past data and it will try to improve itself by using the data at the back end so we'll see what is supervised learning so like what's the difference between you can see that these can be anonymous supervised and unsupervised so in supervised learning you have the prediction that you want to do so you have that target variable with you given what you want to predict so a very famous example I would like to imagine you again imagine you want to buy a house in your locality so what all things you will see as a buyer as a buyer you want to purchase a new house just a simple question what all features what all attributes or what all you can say like buying a house what you will see in that house like what are your expectations so let me describe some and if you have thought about that it's really good so let's discuss some some bigger or you can say there's some major ones the first one is the area of the house that is like biggest thing you want to buy a 2 BHK you want to buy a 3 BHK according to your needs according to the money you have and things like that the first one is the area that you want to buy the area square foot area of the house second is you will see the locality third you can see the like how like how recently it was built is it of a modern style or it is very old built like 10 12 years back or it is just recently built two three years back and things like that okay so how old is your houses third is whether it is located centrally like all the facilities that you need a supermarket a railway station a bus stop and like commuting services a nearby to your house that you will you can see and things like that so those were some bigger level features that we actually see when buying a house so in terms of supervised learning in terms of supervised again it is classifying into two classification and regression so this example that I have taken the first one of house one is of regression time so what how machine learning model will and like first of all I will continue my example so these are the features that you have in your mind so you go to your broker and say I want to buy a 3 BHK and it should be centrally located and it should be new house like max to max two three years old and things like that so based on the information that you provided to your agent or broker the broker will tell you that this house is located in XYZ locality and the price of that house so what is the deciding factor of you buying or not that is price so if that price falls under your budget and you like the house you like the locality you like everything every other thing you will buy that house so that is the target variable that you are taking your decisions you are making your decisions whether to buy or not based on the prices so that is your target variable that your algorithm will predict so that is the case of supervised learning regression problem you want to predict a house price okay based on the information provided so now how algorithm will predict like it doesn't know how algorithm is not a broker algorithm cannot tell you what is the price of the house so based on the past data that you will feed to that algorithm will help the model or help the algorithm to build the model on those data and in future points it can predict the house prices so let's say you have in your area or in that area you have 100 houses you have the area of that house you know when it is built and you know the price of the house you know the target variable you know the price thing so you have 100 rows you give it to model that a machine learning model see this is my data this is my features and this is my target price that is coming out learn it okay you are telling a machine learning model to learn it that is what we call machine learning machine is learning the patterns of your past data and now once the machine has learned you have evaluated your model with good performance a you see an accuracy now you can predict the like predict the price of a new house so you just have to feed in your area where it is located and when it was built like how old is your house and it will predict your house so that is the prediction model and which is a regression one second example of is of classification this is a little bit of different in which we are not predicting any continuous variable that is price but either we are predicting a binary classification or multiple classification so when we say binary classification it is just two okay so it various example of it is whether the person will default alone or not whether this picture of a cell or some tissue in your body is cancerous or not so that is binary either yes or no and if it is a multi classification you can have like in terms of sentiment classification you can be either happy you can be sad you can be neutral you can be joyful so that is like multi classification so as we have mentioned in our examples the diagnostics that I have taken example of cancerous tissue you have customer attention you have image classification whether the image is dog or not simple yes or no whether the image you fed to the model is a dog or not and identity fraud detection whether the person will default alone or not so that is one example of banking one of health care one of marketing one of images and things like that so we have covered the supervised learning in full details and you can see in our regression also we have few examples like you are predicting the advertising popularity you are predicting the weather forecast nowadays we have the weather forecast we tells us the temperature of full one week what will be the temperature in next week so that is we are predicting market forecast you are forecasting the market what will be the response of your customers and the market how it will react estimating life expectancy you have stock market prediction which is also a regression problem you are predicting the price of some stock you are buying at rupees 100 but you are predicting it can go up to 150 so your machine learning model you can predict a house price a stock price using the past date of that stock or the performance of stock and what not so all those are features that has been set or sent to the model and it is predicting out hi and welcome back to the course so the second classification after seeing the supervised learning we have the unsupervised learning and as we can see that these are synonymous and antonyms to each other so we know that the difference is the target variable is missing here so we don't have any target variables in this kind of algorithms so in unsupervised we have everything but we don't have the target variable like in our previous case we have the house prices as our target variable which we want to predict but in these cases we don't have any target variable so as an example the recommendation systems these days we find all these systems in at amazon at netflix the movies you watch and it will recommend based on the data they have so how they will recommend so they will match your behavior they will match the previous data that if a user has bought item a so they will see that that is like a whole different market basket analysis that they do so they see that in their previous records with an item a mostly people bought item b so it will recommend you that may you like an item b so that you can buy so it is trying to cluster you into the same bucket or you can say into the same bunch of people whose behavior is similar to you so let's say on the netflix you mostly watch romantic movies and the comedy movies so the algorithm will predict that what kind of a behavior you have what kind of movies you watch and based on your interest they will recommend you the next best movies that you can watch so that is the power of unsupervised learning and we call it as a clustering so they can target that can be used for target marketing so based on the behavior matching they will target the group of audience using some similar activities and of course customer segmentation they can segment whether these customers are our happy customers these customers are our medium customers who bought like little less and these are our like really bad customers i can say really low profile customers who rarely come to our site and purchase an item so there is classification of segmentation of the customers the second one we have an application in dimension reduction where we reduce the dimensions of the big data or the data which has like various features so you can just understand that if our data has 2000 features 5000 features attributes you will see which are correlated which are not relevant for the application you're solving and you will reduce the dimensionality using your unsupervised approach the third one is very interesting one the reinforcement learning and which we have taken an example of a child this is very similar to that a child is learning his from its behavior or from its experiences so a child know once he put his hand in the fire that this thing is not a good one this gives that this thing or this activity can give a pain so that is the experience that has been stored in his mind and he will refrain himself from doing that activity again similarly the reinforcement learning is kind of a same behavior perform the data into the same behavior so it will learn according to what the steps are there and it will have the penalty associated and the rewards associated so if it is a good thing or a good performance if it has increased some something has been done some activity is done and it has been a good thing so it will reward it but if it turn out to be a bad thing it will penalize so that algorithm is learning what is good what is not a very famous example a chess game or go player game that has beaten the human in in the game of chess there is a there are some other games so that algorithm has beaten the human in terms of playing so algorithm learns by like reiterating different different steps that whether I move this step so what I'm losing if I do this step what I'm gaining so a machine can store all those patterns inside that so that is like they have defeated a human in a chess competition the algorithm has defeated that so that is what we have reinforcement learning it will learn itself it will penalize if the move is wrong it will reward if the move is great or giving some great output so this is the product categorization of machine learning next next let's see our next slide hi I welcome you back to this course so the next thing after seeing the product categorization we have some examples some very fascinating examples so I really want to discuss the first one the thing is there has been a study done by Forbes and some greater you can say the supermarket giants such as Walmart so they have seen a unique pattern in terms of their selling items so in terms of data science we have something called market basket analysis which it does it they will place certain items in such a way that has a high probability that they will sold together so you can let's for example think of a bread and butter okay this is a usual breakfast in our Indian families where we if we are buying a bread or if you are buying a butter there is a high probability that you will buy a bread too so that you can have a bread butter similarly if you have a cornflakes it's a very high probability that you will buy a milk too because you will eat the cornflakes with the milk so that is complimenting each other so these products are complimenting each other and there is a probability high probability that these two will sold together so you can place those items in a shelf like nearby to each other but there is some very unique pattern had they observed like it cannot be thought in our minds like it's completely based on the data so what they observe that whenever a father is coming to buy a diaper for his baby or a male is coming to buy a diaper for his baby they have observed that in a cart they always have a beer a cart of beer in that so what they did they did analysis on you can say past 1 lakh records 1 lakh transactional records that they have so these giants produces bill when you check out from D Mart from your big buzzers so they will give you a bill they have your information what items you have bought so they studied all of them and they found that in mostly 90 to 95 percent they have found the association in terms of market basket analysis we call it as association between these two items so they found that whenever a male father is coming to buy a diaper they find that a carton of beer is always be there in the cart so they have produced this as a cross sell options if a father is coming to buy a diaper he will buy a beer so they have placed their items in such a way that it is boosting their sales so that is like really unique pattern that one cannot even think of but using data it can be proved the second is when we go to supermarket or when we are buying at amazon we see that a person buying x item is recommended item y that is complementing to that item so let's example if you buy a DSLR camera you will see in the recommendation either a tripod or either a memory card because these two products are complementing your DSLR camera and which you might require in your future to click the photos you must be needing your memory options memory card and things like that the third example is your netflix recommendation you watch certain kind of movies and it will recommend based on your experience or based on your interest what you would like to watch next so now this is the person like I would have hidden his name first of all I would have asked you that this person is called the father of deep learning and his name is Jeffrey Hinton so this person what is working on this deep learning concept from you can say several several years ago and he has developed certain algorithms in deep learning and he was a researcher at Google earlier so he's called as a father of deep learning who has really made a breakthrough in this era so we will start now a discussion on deep learning stuff how deep learning is a little bit of different from the machine learning so how does deep learning has been evolved so this is the picture I have included in 1956 this is the size of a hard disk yes this bigger size is a hard disk and to tell you the fact it is just one mb of space so you can imagine that one mb of space in 1956 was this much bigger then coming or we gradually progressed to the 1980 that is 10 megabyte of hard disk and see the cost the cost is around 3000 for 95 dollars and today or we can say the modern era 2000 century we have a 256 GB memory card that like very small memory card of 256 GBs and the cost is just 149 dollars so that is how we have progressed and in terms of science we call it as Moore laws which is telling you the amount of transistors that are keep on adding into the chips and we are evolving from the like past 1956 to 2000s and how does this take a shape and because of this paradigm shift only the deep learning is happening because we need a high resource of computing to perform a deep learning so that is now possible because we can have extensive hard disk we can have GPUs we can have CPU RAMs and things like that which is all based on memory devices so now is it possible to have this deep learning algorithms really working hi I welcome you back to the course and now we will see in this slide the foundation of deep learning we have already covered what is machine learning we have seen some examples when very fascinating examples how data can derive such patterns and now we are jumping towards in deep learning how deep learning is different from machine learning so in terms of deep learning this is nothing but a mimicry of your brain it just a mimic version of your brain how does a human brain or any species brain perform so we have a foundation of deep learning that is the neurons that is what the name derived neural networks which are the heart of deep learning so neurons are also called neurons or nerve cell are the fundamental units of brain and nervous system and the cell responsible for receiving sensory input from the external world and sending a motor commands to our muscles and for transforming and relaying the electrical signals that is like really understood biological concept that we have in our brain so we have several of several neurons in our brain that is responsible for communicating that is responsible for understanding and taking actions based on the information so then writes neurons and axon we have in terms of neurons in a human brain so neuron is the middle part which stores the information and then writes are the you can say receivers and axon is a transmitter so these are like complementing to each other and they are like receptors and they transmit the information they receive the information and the cycle keeps on working now let us see how the deep learning neurons looks like so these are our like these neural networks is nothing but a black box kind of a thing for you but these are generally mimicked as a human brain so as neurons are connected inside our brain so these neurons or this neural network is connected to each other and we have the input layer which is nothing but the features we have hidden layers and we get the output of that so let's take our previous example of predicting a house price so let's assume that your feature one is area of the house your second feature is how old your house is like is it five years old ten years old fifteen years old and things like that and the three is whether it is centrally located centrally located with the market and just like that so just we have three features and based on hidden layers we will predicting the house price so now you can imagine that you can have a different different values for all the features let's say you have 250 square meter of area you have two years old house and you have centrally located yes so that is how fed into the network it will go through all the data all through the combinations so these are like all combination it will go through based on the features that we have provided and it will predict the house price so why deep learning is happening or why it has come into existence when machine learning was there so it is observed that when we use these intensive networks like these hidden layers it is just not one layer we can have multiple layers we can have 20 layers 30 layers and when you see that all the combinations of different different weights that has been doing in this deep learning stuff so you get a better prediction so the models accuracy has been increased when we use deep learning as compared to machine learning so that is the main reason why deep learning came into existence so as you also know that the dominant feature in deciding the price of your house is area only so based on area only it will be a deciding factor so the dominant one is feature one less dominant will be feature two and more than less will be feature three but that is on our you can say our human instinct so what does model will learn that will tell you the real thing so we know that feature one is the most powerful one so you can calculate the feature importance and get to know that feature one is driving the sales or your prediction of the house so that is the power of deep learning that can try different combinations of weights and it will come to the optimal stage by doing back like these are the we will see in detail when we will do the course of this deep learning and machine learning but this is just an introduction to them so we will have various algorithm that will set an optimal value for these weights and come up with the best possible iteration or best possible weights for these features and we will get our output based on that so to name them we have an algorithm called back propagation we have stochastic gradient descent stochastic gradient descent that will perform and get us the optimal result for this neural network so now coming to one example of deep learning and this is like really happening in this world in this unprecedented times which we all are seeing the covid the pandemic that has been rolled out so just to imagine how deep learning is contributing in the research of the pandemic or how it is helping the researcher the medical professionals is a live example which I have took of this data and we will be solving this problem in our deep learning course this is like very very interesting problem of detecting whether a person has a covid or not like this is a quick test and the most accurate one like most I will not say most but better accurate one and to get the results at very early stage so this data is available on Kaggle so Kaggle provides you free or open source data so it's a competition that they have organized and these are real images all real images they have provided of chest x-ray scans that these healthcare institutions have done and they have made public so that we can learn or a model we can build a model to help the healthcare professionals and doctors around the world so these are real images of the real person who have undergone the chest x-rays while being suffering from a covid so the photos or the inputs we have is a person is either having a pneumonia or a covid positive or a normal thing so a model is trained on these three images so model will learn the patterns of these three how the how the person with the covid looks like how the person with pneumonia looks like and how the person chest x-ray image of a normal person looks like so now imagine the model is built on you can say 10 millions of record and we will completely build this model from scratch in our deep learning or machine learning course so in our machine learning course we will see an introduction to deep learning and we will do this one example in that so that covers from scratch from collecting the data from data preprocessing from feature mapping and then building a model and I worked on this data and achieved an accuracy of 95 percent so your model is 95 percent times right that a person is either covid or not so you will use cnn cnn is nothing but your convolution neural networks and this will help you to distinguish whether a person chest x-ray is a covid or a normal person so now coming to our ending slides steps or you can say the building blocks of machine learning or deep learning both will look like similar thing or both will look like these steps so this is very simple process that keeps you worrying about what is ml what is deep learning just learn or just like feed in your brain these seven steps if someone says how machine learning works or machine learning model deep learning model works just reiterate these seven steps first of all you will gather the data you will prepare that data in terms of feeding to algorithm you will preprocess you will clean that data you will build to feature engineering then you will choose a model which you want to build you either want to build a tree based model decision tree random forest xg boost or whether you want to do a deep learning neural network you want to build computer convolution neural network you want to do computer vision so you have to choose a model then you have to do training you will train the model then you will evaluate your model how is it performing and if the results are somewhat varying you can tune the parameters so hyper parameters are nothing but your model specification or your training specification things which you can tune something to improve the results so then you hyper parameter tuning you will improve your model and at last you will predict predict what you want to predict you want to predict house price you want to predict cancerous cell you want to predict loan defaults and the use cases are endless so that is the application of machine learning and deep learning you can apply it to any business domain and use case and we have covered it a little bit we have seen the difference between machine learning and deep learning and we know that based on deep learning it is quite more extensive it takes long time as we have several combinations to try out and it takes really really long so the pre-trained model that google and facebook has released they took them months to develop so once you run the model it will take months to pre-process based on the data you have so they feed millions of millions of records to build up very good models so as you can see that requires larger data and machine learning require less data and it is a rule based and data driven or you can say tree based but that is your neural networks are black box so they are based on your permutations and combinations and different combinations of weights so training time is short and here it is long so that is machine learning is good for classification task and clustering so and when you want to do something with images video audio and text it is deep learning deep learning is widely adopted on all kind of unstructured data image video and textual audio so you will prefer the deep learning algorithms and you can train on cpu in machine learning it will take less time and in terms of deep learning you will require your GPUs as well for training it you can say in less time so that sums up the whole this course this whole course what is data science and what all subject or the things that you need to know so this is the Venn diagram of data science you need to know machine learning if you have traditional software programming knowledge that is really good you will need to learn a little bit of maths and stats you have some knowledge of computer science and the subject matter expertise so that is really matters when you are working for some business domain if you are working for business BFSI that is banking finance you should have a knowledge of how banking and finance and this insurance institution works if you are working with healthcare you should know how this particular disease works and how to get distinguished and things like that so that is subject matter expertise so in this boot camp we have done our first introduction to data science and the next course is that we will be coming up with is complete python boot camp that will start from very basic of the pythons and covers each and every concept of python that is your oops concept your classes your files your functions your algorithms that are built in python and we will do one capstone projects also so third one is your business statistics that will come with your statistics knowledge and your you can say hypothesis testing your mean median how the normal curves look like and whatnot the other other course we have the fourth one is SQL SQL and that is really important when you are working in the industry you have to prepare the data before building a machine learning model so that is really a good skill you should know for becoming or getting your hands dirty in the data science and at last we will see the machine learning in which you will learn how to use that data and use that information and build a model out of that