How to Become a Data Scientist (Learning Path and Skill Sets Needed)

Key Takeaways

welcome back to the data professor YouTube channel if you new here my name is Tim then not a cinema and I'm an associate professor of bioinformatics on this YouTube channel we cover about data science concepts and practical tutorials so if you're into this kind of content please consider subscribing so in the last two weeks I have released a video which covers about the overview of machine learning model development which was inspired by one of the infographic that I have drawn and shared on social media such as Facebook and LinkedIn as well as Twitter and collectively it has received more than 2000 likes and so I figured out that it is probably an interesting topic that I should create a video about and which I did on the development of machine learning models so if you haven't watched that video yet please find the link down below in the description and so two days ago I released another infographic which covers the landscape of data science so the inspiration of that infographic are coming from you guys from various channels such as on Twitter on reddit on Facebook so the question goes what is the pathway on becoming a data scientist or what are the skill sets that are required to become a data scientist what kind of courses should you take to become a data scientist and with the explosion of the field of data science with the introduction and emergence of new data analytics machine learning frameworks which might leave the beginner behind because the field is moving very rapidly and it might be a bit difficult to keep track of that and so I spent the couple of weeks doodling on the iPad and infographic which summarizes some of the key concepts frameworks skill sets and important data science concepts and topics that you should consider when first starting out or to keep up to date on the field so let me open up this infographic that I was talking about so as you can see here on linked-in on my own LinkedIn profile I have shared this data science landscape and so here's a look at the Facebook post of this infographic and yes indeed this infographic is meant to be a concise summary or a verse I feel of the field of data science the landscape of data science so please feel free to use this as a rough guideline on some of the topics that you should learn about or be aware of so to download a copy of this infographic on the landscape of data science please find the link in the description of this video okay so let's have a look so in this infographic we have a total of eight major concepts here data pre-processing statistics mathematics software engineering data visualization machine learning and soft skills so a disclaimer before we begin so this infographic is not meant to be an exhaustive list of all data science concepts because if it were it wouldn't fit into this one page but this is a modest attempt to provide a summary verse I feel of the landscape of data science so let me know in the comments which topics that you would like to see and if I can incorporate it into this infographic or perhaps create a new infographic let me know or if you have some ideas on a possible infographic that you would like me to create let me know also down below in the comments so greatly appreciate your ideas okay so let's have a look at the first concept in order of the workflow of creating a data mining or data science model so one of the first step you would have to be aware of is about the data and so the first step would be to obtain the data that you are going to use for your analysis and so data can be structured or unstructured and if it is structured it is normally in the form of a tabular format in which there are rows and columns where columns will describe the variables and the rows will represent the data samples in the dataset and for the columns or variables most will be the independent variables for the variables that serve as the input and some will be the output variable or the variables that you would like to predict the outcome of which was served so to say s8 class label and so as one of you guys have suggested natural language processing is missing in this infographic and so I'm thinking of probably having it included in obtaining data so in natural language processing the input data would be the text and so these texts are essentially in unstructured form and other unstructured data could include as well the audio the image the video either it will be pre-recorded videos and real-time videos coming from computer vision and so that's obtaining the data an important part of the data pre-processing workflow would be to handle missing data and so one of the previous infographic that I have created is dedicated to how we can handle missing theta and the most easiest way to handle the missing data is to make the dataset complete meaning that columns or rows that contain missing values will be deleted from the data set but this counts at a cost of a reduction of the data size so the number of columns or the number of rolls could be significantly trimmed down however there are ways to replace these missing values either by using the column mean median or mode or to predict the missing value in the context of other values that are already present in the column okay and another important concept in the a pre-processing is data cleaning so this is to ensure that the text for the strings or the numerical values are properly and correctly spelled out or that it does not contain any type of errors or also to maintain the consistency of the data set such as the naming of the data values inside each of the column because if the values are misspelled then this would give rise to in the categorical value inside the column so great care has to be taken in that step also another important concept in data pre-processing is feature engineering so aside from obtaining the features image it could come from databases or it could be freshly measured and an interesting area in this would be to find out ways on how you can engineer novel features which could come from simply subjecting it to logarithmic transformation combining multiple variables together simple addition subtraction multiplication or division finding the ratio of variable day to be the variable C to D or even multiplying it by some constant values so these would allow you to generate novel features however you have to be aware of the meaning of such generated features and how that will potentially be interpreted after the model has been built during the feature of importance or model interpretation phase and another important topic and the data pre-processing concept would be feature selection because because the feature generation and feature engineering step could come up with several thousands of variables and perhaps many of these will contain no values at all or there will be inherently collinearity in which several variables will contain the same information and so we will have to handle such large volume of features by performing feature selection so feature selection could be done by for example removing variables which contain very low variance because if variables contain low variance it means that it does not provide any meaningful information so for example if 99% or 99.99% of the values of a particular column or variable contains the same value for sample a value of zero and point zero zero zero one percent contains a value such as one and so for the purpose of developing robust models such variables would have to be removed or another would be to remove variables which exhibit similar trends and behavior by computing the intra correlation in which it is essentially a pairwise Pearson's correlation coefficient matrix and so for a given pair of variables which has high correlation coefficient value we will remove one of them and keep one of them and so this will be performed iteratively until we obtain a set of variables which contains the Pearson's correlation coefficient value less than the establish direct short value so for example we could set the threshold value to be zero point six or seven point seven and if a pairwise between there were one or blue to contains coefficient of greater than zero point six which is threshold then we would remove one of them right and then this is the first iteration and it do the same thing over and over again until there is no pair which exceeded two I showed that we have established and so this is roughly concludes the data pre-processing group of concept and let's hop on to the next one statistics so undeniably statistics is an essential part of their science and it is at the backbone of data science and some of the core concepts of statistics would include informational statistics hypothesis testing experimental design and descriptive statistics so for example in descriptive statistics we are able to get a glimpse of the relative distribution of the data the comparison of multiple variables by means of comparing the mean of the variables evaluating differences between variables either two variables as in a t-test or amongst multiple variables snv anova okay so let's hop on to the next concept which is mathematics so at the other end of the spectrum mathematics provides the fundamentals in which it will help you to understand the underlying mathematics behind several machine learning algorithms such as speed learning new network principle component analysis etc and so some of the concepts here would be linear algebra discrete mathematics optimization ability theory calculus real analysis geometry so as you can see mathematics and statistics will help able to understand the logic the concepts of the learning algorithms under the hood which will also help you to understand the limitations the strengths and weaknesses of different learning algorithms which would help you to select the optimal learning algorithm or select the optimal statistical test to evaluate your hypothesis validate your hypothesis as well as in the development of your own machine learning model and so another important concept here would also be data visualization so in data visualization it is essentially the creation of graphical plots to visualize the distribution of the data points as well as the composition of the data the relationship between variables between data points so each of these subtypes here comparison relationship distribution composition will be sub branching into the different types of plots okay so now let's hop on to the meat of the data science landscape which is machine learning and so this is a very important concept here and it might be mistaken by newcomers to the field in which they would focus on only using machine learning or they might understand that machine learning is the only important concept that they should be aware of but in fact this is only the tip of the iceberg and so there are a lot of stats of essential concepts such as statistics data visualization that thematic sphere pre processing software engineering as well as programming and also soft skills so most of the attention in the field of data science might be given to machine learning and perhaps programming as well which I will cover into some moments so machine learning might be the attention grabber of the field of data science as it represents artificial intelligence learning from data making sense of data using fancy algorithms deep learning support vector machine but under the hood learning wouldn't be robust if the underlying data is played with missing data missing values features are improperly calculated data is not properly cleaned in appropriate statistics are used to evaluate the data set right so in order to develop meaningful machine learning models one would also have to strengthen their background on the basic concepts as mentioned previously right so programming is essential player which modulates statistics their pre-processing mathematics data visualization and machine learning and there are a lot of programming languages up there and so at the fundamental level you would want to learn either R or Python for your data analytics right as I have just released a video about which programming language should you learn for data science and in that video I covered about R and Python and of course there are several other languages that are up and coming but historically are in Python are in the game longer and so as a result it has a lot of accompanying libraries and packages that are available for making many of the data science tasks a lot easier particularly if you're working in the economics or life science biology chemistry there are already existing packages that will make your analysis a lot easier going to the specialized function which would otherwise require you to create your own function which could be quite complex so of these languages if I could recommend it would be nice if you could use - - if you're working in a Linux environment or UNIX environments such as on a Mac and Linux as well and lately Windows also has an application in which you could run the command prompt from Ubuntu and of the or and Python languages I would recommend you to select one of them and used for performing the various tasks of data science such as pre-processing the data performing statistical analysis visualizing the data performing some mathematical probations as well as something the machine body model and if you're working with big data set then SQL is an indispensable tool so you should also learn that as well so aside from programming it would also be useful if you could also be aware of some of the subcategories of this concept of software engineering so if you're working in a big team you might already have someone working as a bigger engineer to assist you in tasks such as model deployment parallel computing optimizing your code to make it run faster or performing version control performing code optimization as well as debugging or testing of the code also if you could also read up on the best practices for software development be aware of data structure and web development particularly if you would like to deploy your model so that it will be accessible online or via the intranet so the last concept here is the soft skills so the affirmation seven concepts would be the technical skills of data science and the soft skill of data science should not be overlooked and so the soft skill will essentially be those that allow you to interact with other members of your team or the stakeholders or your customers or a different department within your company such as you would interact with people from the sales department people from the marketing department providing them with insights from your prediction model and you could also learn about the data or the interpretation of the thing by talking to the people from various departments of the company so they provide you with domain knowledge an important skill would be to communicate your data right so storytelling in the form of data visualization so how can you make a beautiful appealing and meaningful data visualization which will be essentially data storytelling and if you could communicate that to the stakeholders so presentation skill would be an important thing to have writing skill problem solving creativity and also grids so greatest problem one of the important traits for a beginner who is from a non-technical background so oftentimes learning a new discipline such as data science is an overwhelming and deeper so without grits you might give up in the first couple of months because of the unfamiliarity of the field because of the overwhelming concepts that you will have to digest in essentially climbing up the mountain of various data science concepts so the first moment would probably be programming and then mathematics also and the various algorithms to learn or choose or select in the development of your machine learning models so grit perseverance is a must if you are starting out from a non-technical background and in my opinion one of the most important would also be curiosity so we carry also the comes the origin of your urge or desire to know to learn to make sense of the data so you are kind of like a news reporter you want to go behind the scene you want to get the data you want to understand the root of the costs of the problem that you are going to analyze so in that you will have to do many things aside from building models you might have to talk to the stakeholders you might have to read up on books in the domain so that you could acquire domain knowledge so curiosity will spark your motivation your urge a desire to move forward in your data science project and as a result data science is a lifetime learning and either because new algorithms new software packages will be introduced and the field will evolve and so having an open mindset honing your skills learning new skills is crucial for success in data science and so I wish you best of luck in your journey into this very exciting field data science and so if you're venturing into this field please have a look at the infographic think of it as a blueprint or a starter map which will help you to explore what is there to know indeed of some so if you find value in this video please give it a thumbs up and comments down below sharing us your journey into the Guv science and also if you would like to see a new infographic what topic would you like to see comments down below thank you for watching please like subscribe and share and I'll see you in the next one but in the meantime please check out these videos