Take Your Data Science Projects From Good to Great

Key Takeaways

hello everyone that can hear back with another video for you today I'm talking about how you can take your data science projects from good to great now in my most recent video I talked about the types of data science projects you can do to effectively learn the field and to catch the eyes of companies I'm expanding on that a little bit and focusing on the areas in which you can really improve and make the greatest returns if you put an effort so these are a couple of the kind of key steps to having a cohesive data science project that will really stand out from others that are applying or that you're showing this to as usual if you enjoy this video please hit that like button and if you want to see other videos similar to this please subscribe and hit that notifications to be alerted when I post a new video now something I think is really important when you're starting off a data science project is having a clear vision about what the project is and what you want to find out the ability to tell a story about your data science project I think is one of the most important skills that you can have especially if you're explaining it to a potential employer if you can relate your personal interest in the topic you can relate the company you're applying to to this specific project it's always going to come off better you're going to be more of a motivated to work on it and you're also going to be able to say the exact type of value that you set out to create and the exact type of value that you did create the next thing that I think can really set a good data science project apart is attention to detail with the data and focusing on data collection so it's always interesting to me when someone writes a web scraper or gets data from a unique source so let's say I even collected my own fitness and eating data and did a project on that that to me is going the extra effort and showing that you're really invested in data science and and integrating it with your life I always look for people that are passionate about the field and you're willing to take a couple extra steps that are related to data data collection not necessarily data science I think that that shows well on a candidate it also says something about your personality and your character that you're willing to go maybe an extra set that other people don't it's great to get data from kaggle or from some of these other sources but I'm always again impressed when someone goes out and it gets this data themselves the next thing that I like to see in data science projects is creative feature engineering so you have a specific data set I either like to see you go out and get another data set and append of that so let's say I have information on a specific group of people if I have their zip codes perhaps I can append some information related to their incomes and that can add new information to my analysis also again a fives if codes and let's say I want to estimate if they're going to attend a specific school I can actually use that zip code to feature engineer how far they are from a specific school and use that as a feature as well so that is something that I think just like the the previous idea is going the extra mile but this one can have perhaps even more of an impact on your model I like to see how creative people can get here because this is one of the areas in data science where you can be the most creative and you can think the most outside of the box so definitely explore different feature engineering technique techniques that goes as far as looking also into principal component analysis or using some sort of clustering in your inputs a PCA is great especially if you have a very large feature set but it can also help you understand which features are related to each other so keep those tools in your tool kit and make sure you use them or at least consider using them in your projects I really like to also see people get creative with the algorithms that they use in their data science projects so a lot of people just go through the gambit they try four or five different algorithm algorithms see which one works the best well you can also start looking into ensemble approaches or you know different ways to tune your models and that is what it takes to go the next level in my opinion so it's not just that a random force is going to be the best for this but a random forest combined with a multiple linear regression might actually generalize better so layering these models understanding what some of the you know deficits are in specific models and seeing if a combination of a couple different can improve your accuracy or improve whatever you know variable that your you're wanting top demas so again I encourage exploring the combinations of models not just models by themselves after you have a completed model it's always nice to see it packaged and usable to you know anyone and a company or through a website so I like to see it when people productionize their project model so that's could be as simple as putting your model into a flask wrapper and making it an API endpoint through your website so let's say I made a blood pressure predictor and it took in someone's basic weight their height and a couple other features I could put that on my website and you as a consumer could go in and put in your variables and it would return what your systolic and whatever the other type of blood pressure is that was projected from the model that to me is a cool implementation use case it's something that would probably be used by people on your website and that's also something that's very common in industry where you're actually making your model useful to someone that shows that you can go from square one of collecting the data all the way through to production ization and that's a great skill that frankly not all data scientists have and that's a very useful skill in my opinion the next thing that I think is important getting away from the actual model building or any of the technical components is creating a model that is valuable to someone so for example if I make a trading model and it can actually make me money that model is in you know inherently valuable to me and it could be valuable to other people in you know the generation of wealth if I create a model that helps you know high school students choose what college they should go to that helps other people as well and that is the goal of data science is to help other people help a corporation make more money help you make better decisions help other people in general and if your model serves to do that and people will actually use it that is the sign of a really really good project and the Sun again of a great project is if you can actually get someone else to use this model of yours I really recommend people go to nonprofits or their school or someone that might have a problem that you could help solve because that adds purpose to your project if you can create a project that will help one of these organizations then that is about as good as something can look on your resume well you heard the sirens I think that that means the police think my video is going to long so let's end it here as usual thank you so much for watching please in the comment section below write it write what you think makes a great data science project good luck on your data science journey