How To Implement A Real World Usecase or Project For Data Science

Key Takeaways

hello all my name is Krishna and welcome to my youtube channel in this video we'll be discussing how you can implement a real-world use case and data science by using machine learning or deep learning so I'll be discussing about some of the techniques now why this question is very very important guys because many of you from many positives are asking this particular question that so Krish we have actually finished our online course we know how to implement the machine learning and deep learning projects but now we need to have some real-world user expediting real-world experience by solving some real-world use cases so this was the question by asked by many of my subscribers so I want to basically tell you that how you can basically create a very good machine learning or deep learning projects and what are the steps that you will basically follow so the first step is that guys you have to learn machine learning and deep learning ok you have to learn machine learning and deep learning properly so that you are ready for implementing the machine learning or deep learning projects now the next thing is that how do you find a real-world use case now one way to find a real-world use case guys you can basically go to kaggle ok because in Kaggle every month there will be some competition that will be put up over there and that particular problem statement is of a specific company ok is of a specific company so that is one of the way how you can basically get a very good data set to solve a particular problem statement the other Prag the other thing is that what you can refer right now the main thing is that in order to solve any use case you have to get some data right you have to get some huge amount of data so Kaggle is one source where you can get that particular data you can get the problem statement the other data source or repository that I'll say is that there's a government website even in India and outside India in US where they will be giving the data set for absolutely for free so by taking that particular data set you can create a use case and you can start solving it now when I say you are solving that particular use case and when I talk about that particular URL guys I will be giving that in the description box you can basically see that particular URL and you can get any amount of data that you want with respect to specific domains now why I am telling you to refer Kaggle because skagle has datasets from different different domain you will find datasets with respect to an e-commerce site with with respect to retail domain with respect to a financial domain with respect to healthcare domain and so many domains are there so they will basically be having a lot of problem statement you can pick up what is your interest in which domain your interest is you can pick up that particular data certificate start swallowing okay now you have done that now when you are solving the data set dies apart from machine learning and deep learning make sure that you also learn flask or Django okay so flask or Django is very very important so this is my third point that you have to learn flask and Django because at the end of the day you want to create a web application where you will be integrating your model okay your model can be hosted as an API okay so you have to know how to implement how to implement a web application it it may be a basic web application initially because I want to create an application from end to end okay from the data gathering stage till the complete data deployment and this should be all together in one particular web form or web application or a software something like that okay so this will actually give you like this is how industries also it this this is how every project is basically implemented so I want you all to also create a web application for that please learn flask and Zengo tutorial and if you want any reference part for that particular site from where you can learn flask as mango Cory Schaffer Cody Schaffer is the YouTube channel used the poser name he basically uploads free videos in YouTube please go and refer that by that you will be very very good at flask and mango because he uploads wonderful tutorials okay so now you have understood that you have to learn machine learning and deep learning you have to for the data set for the problem statement you can refer Kaggle and some of the government repositories there where they are dumping all the data for free of cost you can download it and can solve any problem that you want the third thing is that you have to learn flask and Zango okay and once you're learning flask and angle you will also have to learn some deployment techniques now initially what you can do is that in order to learn any deployment techniques you can go ahead with Heroku because here a couplet form provides free deployment for some some model size okay so you can use that and if you are creating a machine learning problem statement the model size usually will be slop small you can basically deploy that and you can actually use that particular free service okay you'll not get a live web URL and if you don't know how to do the deployment and haricot guys you can refer my deployment playlist in my youtube channel itself okay so I've shown you how to deploy an NLP model and deep learning model using Heroku okay you can basically use that particular stuff and it is free of course you don't have to pay anything so so the deployment technique after you learn Heroku you can also go ahead with AWS and assure you'll be getting some free dollars credit where you can basically try out different different deployment techniques like AWS ec2 instance and many more things so now observer I'm covering everything from end to end I know machine learning and deep learning I basically have the data set I have the problem statement now I know flask and then go where I'll create the front-end application and from that particular friend and application it will interact with the model which is basically deployed in the Heroku platform or some other platforms so all this particular steps but now creates a package or web application which is just like one product right now the next thing is that what you'll be creating in your web application that is the main thing now understand guys whenever you are picking some data set from the kaggle write that data set will be having some number of features right so you should try to come up with a story with a plan like how you are going to basically use that how you are basically going to take the features from your web application okay so that you have to observe over here how you are going to take it so you can basically create 10 to 11 fields just like a application form where you're just filling the forms you are getting that particular data and then you create a separate pipeline they should be two pipelines for these guys one pipeline is how a model is getting trained the other pipeline is that whenever you are taking the new data then you are again doing that feature engineering feature selection and then predicting by giving that particular data to their model and getting the output okay once you get the output that you can do anything with that particular output and this is how you have to do it is just not a simple thing guys I know there is a lot of work to be done so it is always better that you make you take your friends take three to four people in your group and start implementing this one by one and I will not say that in the first day only you will become perfect it will take time to solve a real-world use case now I have a candle problem statement I want this in two in a complete web application form so basically I want to integrate that in a web application and continuously take a data and find out the predicted output and this is how we should basically do you can also use a database so that you can basically you know whenever you are getting an output you can store that value in a database and periodically you can check that output how good it is how bad it is and based on that what you can do is that you can additionally train your model again and again if the accuracy is going down so this is how you should go ahead and creating a real-world use case and guys Kaggle is the best source I'll tell you because toggle problem statements has a real world problem statement based on the industry standards because industries are only giving the problem statements right so you you will be seeing lot of financial problem statements are their retail commerce traveling domain problem statement insurance domain problem statement and the whole lot of health check health care sector so if you see over there how to how to predict dingoo how to predict malaria cancer diseases you know so all those things are there already there was those many problem statements are there you have that particular data set right so all you have to do is that try to make this into a whole web application or a whole product wherein you are doing all the steps in a pipeline and when I say pipeline guys remember your training phase of your model should be in a separate pipeline okay your your your I mean your application when it is taking the new data and doing the other step that is your feature engineering feature selection finding out the output that should be in on another pipeline okay so try to make this two and after these guys made sure that whenever you are doing this things write blogs regarding it so that you can document all these particular steps what you have actually done make that documentation write a blog regarding it create a github profile and properly just note down all the steps properly over there right and when you are writing that particular documentation tell that which algorithm has worked how many algorithms did you try right and in that particular algorithms also you can say the disadvantage what was happening with respect to that particular data whether the accuracy was decreasing or increasing why did you go and select the other particular algorithm so all these particular steps has to be noted down over there and you can also create a github profile very nicely structured or you know write the content write the reference paper you are basically using any research paper suppose you have used to solve that particular problem please give that particular reference you know and that will also create a habit for you all to read research papers which is very very important right so this will all together give you a very good knowledge and like this even though you are a fresher if you have done this all things obviously the same thing will happen in real-world IT industry right whenever you have a new use case now additional to that if you are looking if you are looking for something new that basically means you have fine-tuned that particular model in such a way that you are able to get a state of heart algorithm then you can also go ahead and write something new right you can also go and write a research paper for that right but make sure that you provide all the references that you are basically used what's fine-tuning you have actually done to get the state of heart algorithm so all the steps has to be clearly jotted down note it down over there and shown it to the and you can also again give it to the community completely for free they'll recognize you they'll definitely recognize you how many followers you'll have in your greater profile you'll only not know you know because this increases exponentially if one person find that particular use case he may share it to someone else and they will be using you using it and definitely recruiter will give a recognition for that because in your resume you'll be having your blog URL you'll be having your github profile once they go and see over there they see the number of followers over there they see what all steps you did what all research papers you have basically given now and based on the your use case they'll go and read it how you've solved that particular problem statement I think that is must and here the major factor is that you are covering each and every thing that a data scientist - now after that you can also look head for automating things like this data that you are initially capturing from web application then you can automate these things you know continuously you're capturing the data you're getting the output then after getting the output you are storing it in some visualization - using some visualization tools you are creating reports right and publishing that reports in this particular server so that the other stakeholders can see so all these steps can be covered up and made this into a real world sin I have the real world use kiss guys and finally I think if you just follow this steps you will not have to say that so Krish from where should I find out real-world use case right now okay so this is it guys so I hope you understood this particular video please do subscribe the channel if you are not already subscribed I'll see you in the next video have a great day ahead thank you one and all