Speaker 1: Say you get a bunch of separate Excel files from different stores with orders data on a weekly basis And you need to collate this data into a single table for further analysis You could just use Excel, create a new workbook and copy and paste your data onto the same worksheet Or you could just run this one line of Python code. That'll do the exact same. Cool, right? Hey, my name is Mo Qian, and I work as a data and analytics analyst within the financial services industry. Out of the technical skills I've acquired throughout the years, Python was by far the most challenging one for me to learn, which is why in today's video, I'd like to show you how you can learn Python for data analysis efficiently, or at least much more efficient than me. I'd say Python is the most powerful tool in my data analyst skills arsenal as I can clean and transform data with it, create data visualizations, or write scripts to automate certain tasks or processes. Learning to code in Python was not easy at all. I've had my fair share of failures along the way before I could eventually write neat, clean, and efficient code. In this video, I'd like to share with you the biggest mistakes I've made and the lessons I learned from these mistakes which helped me succeed in learning Python for data analysis. I really hope that you can get value out of my learning story, relate to a part or even various parts of it, and accelerate your own journey. So first of all, let me quickly tell you about how I completely failed at learning Python right when I started my career five years ago. I was working as a RISC graduate within the banking industry after I just finished my master's degree in finance and economics. For those of you who don't know much about graduate schemes in the UK, they're jobs designed specifically for university graduates. You usually sign a two to three year contract and get to do a different rotation every six months. I was in my second placement working in a risk modeling team, and this was when I first encountered Python. Just for reference, after graduating I had zero technical skills, no Excel, no Tableau, no SQL, and of course zero Python. I've picked up all of these skills after I started working, so in this placement I was surrounded by very, very smart people. They all had master's degrees and PhDs in quantitative subjects like quantitative finance, statistics, econometrics, and were also super technical. Let's just say that sometimes I didn't even understand their questions, not to mention answer them. I felt pretty disconnected, out of the loop in terms of skills and knowledge, and tried to learn all the Python I could within two to three months so that I could use the rest of the time to apply my skills and make an impact. Safe to say, I was too eager. I didn't spend enough time learning the foundations, blew through the topics quickly, and by the time I got to the more advanced concepts like classes or writing scripts, I was pretty lost. Learning Python was was a challenge that I severely underestimated. Mastering everything in Python is extremely difficult, if at all possible, but I didn't know this back then. I wanted to run a marathon when I wasn't even able to make the 5K mark as I tried to create and run automated credit risk models when I could barely understand a simple class within the code. Being in this team humbled me for life and completely changed the way I look at different levels of technical skills. So now that you know how I failed at learning Python, let me tell you how I actually succeeded in the end. I built on my mistakes and I created a structured roadmap focused on Python for data analysis. And I cannot highlight the emphasis on data analysis here. Learning everything in Python will take you ages, So narrow down your focus by learning the basics very well before moving on to mastering essential libraries like NumPy, Pandas, Matplotlib, and Seaborn. And by learning the basics, I mean build a strong core knowledge of what data types, lists, dictionaries, mutable or immutable objects are. Practice looping, be able to write functions, lambda functions, and other basic built-in functions. Have a basic understanding of what object-oriented programming is by learning about instances and classes. I made the mistake of copy and pasting a bunch of code, thinking only about the end result and getting it done, which was not great from a learning perspective. Try and type out the code yourself, as even though you can easily copy and paste or just ask some AI tools to write some code for you, I feel the code sticks with you much better if you actually type it out. Trust me, being able to actually code from scratch will help you so much when it comes to altering some code that you just copy and pasted or when you need to understand someone else's code and pick up the work from them. After building a strong foundation with the basics, you can move on and master the essential libraries. Let's cover NumPy first. It's used for numerical computations in Python. Its popularity mainly comes from the fact that it supports large multidimensional arrays and matrices and a bunch of math functions that you can use to operate on these arrays efficiently. It also has a broadcasting feature that helps you perform operations between arrays of different shapes and sizes. Say for example, if you have a larger and a smaller array, NumPy automatically replicates the smaller array to match the shape of the larger one. NumPy also integrates well with other libraries such as Pandas or Matplotlib or SciPy or Scikit-learn if you're into machine learning, as it's a foundational library in the Python computational ecosystem. It gives you a seamless workflow for data analysis. Moving on to the Pandas library, which is an open source data manipulation and data analysis library for Python. It's designed to make working with structured data such as tabular data or time series data more convenient and efficient. It has two primary data structures, series, which is a one-dimensional array that can hold any data type, and data frame, which is a two-dimensional data structure where each column can hold a different data type similar to tables in Excel spreadsheets. Pandas simplifies the process of reading and writing data from and to various file formats like CSV, Excel, Parquet, or PQL. Moreover, you can easily manipulate and transform your data using the data cleaning and preparation functions, as well as handle missing values and categorical variables. Pandas also has strong indexing capabilities, allowing you to select, slice, filter data based on your chosen criteria. You also have many easy ways to access specific rows, columns, or even subsets of your data using labels, Boolean expressions, or positional indexing. Pandas comes in really handy when working with time series data, as it has extensive support for working with it. You can use the functionalities to handle time-based indexing, resampling, time-shifting, even rolling window calculations. Very useful when analyzing and manipulating financial and stock market data. Pandas also integrates well with other libraries such as NumPy, Matplotlib, or Scikit-learn if you're into machine learning, giving you a seamless data analysis workflow by combining the Pandas data structures with the computational and visualization capabilities of other libraries. Speaking of data visualization, let's move on to the Matplotlib and the Seaborn libraries. Matplotlib gives you a wide range of tools and functions for creating a variety of visualizations, such as line plots, bar plots, or histograms. And Seaborn is a library that's actually built on top of Matplotlib. Use Matplotlib to create high quality plots with customizable settings for fonts, colors, line styles, or markers. You can modify the axes, labels, titles, or legends, and add elements to your plot as well. The customizability is insane. You can pretty much control every aspect of the visual. You can also create multiple plots within a single figure using subplots, and you can arrange the subplots in a grid or any other custom layout you prefer. Subplots are great when you wanna present multiple visualizations in a single image. Now, Matplotlib is great, but if you wanna go the extra mile and make your visuals even more eye-catching, use Seaborn as it enhances the visual aesthetics of plots compared to the default styles of Matplotlib. It has a set of predefined themes and color palettes that look much more visually appealing and professional looking. Seaborn complements Matplotlib very well, as it simplifies the creation of complex statistical visualizations. You can easily create box plots, violin plots, or regression plots. You can also just as easily visualize categorical data by using scatter plots, count plots, or bar plots. You can then use these visuals to compare groups, display proportions, or highlight relationships within categorical variables. One of my favorite things about Seaborn is the beautiful heat maps you can create, which you can then use to highlight patterns and correlations in large data sets. Both Matplotlib and Seaborn work well with other statistical computational libraries, such as pandas or NumPy, giving you, again, a seamless workflow for data analysis. And that's it. That's the end of the video. If you enjoyed this one, make sure to check out some of my other videos right here. Thank you so so much for watching, and I'll see you in the next one.