Thursday, September 28, 2023
HomeAIData ScienceHow to make a data analysis project you want to read?

How to make a data analysis project you want to read?

Table of contents

  • 8 actionable tips to stand out at work
  • Tip 1: Choose your analytical goals over the tools you want to use
  • Tip 2: Have a methodology ready
  • Tip 3: Brainstorm like you used to in everyday life
  • Tip 4: Do preliminary analysis to identify the most promising narratives
  • Tip 5: Use storyboards to build your narrative
  • Tip 6: It’s about them, not you
  • Tip 7: Get feedback first
  • Tip 8: Pay attention to details
  • Other Helpful Tips

8 actionable tips to stand out at work

The most important part of building a career in your field is getting known for your work. When it comes to data analytics, building an impressive project that showcases your knowledge and expertise trumps all other avenues like certifications and courses .

So how do you build an impressive project? More importantly, what makes an analytical project impressive?

In this article , I’ll share eight tips that helped me become a Kaggle Notebooks Expert by building narratives across various datasets . Let’s get started.

(*Translation Note 1) Mr. Yadhunath, the author of this article, said in the AINOW translated article , “Online courses won’t make you a data scientist “, that qualifications and online course completion certificates are useful when looking for a job as a data scientist. Instead, he emphasizes that the data science project that he voluntarily worked on will be a material that can be appealed .
(*Translation Note 2) Narrative refers to the whole act of telling a story in context by a speaker . The story is what is told in the narrative . While the story is complete and unchanging, the narrative changes according to the situation . A narrative is semantically closer to a presentation than a story. However, narrative and story are often confused.

Tip 1: Choose your analytical goals over the tools you want to use

“Always remember that your narrow focus determines your reality.” – George Lucas

It’s all too easy to get lost in a compelling package of data visualizations that steals our hearts . There’s nothing wrong with learning a new tool, but a certain moderation is required when working on a project .

The ultimate goal of an analytical project is not to show off your knowledge of a new tool, but to discover useful patterns in the data provided. So instead of worrying about which tools to use, it will be more fruitful to focus on asking questions of the data.

Of course, depending on your project’s requirements, you may prefer one library over another . This case is related to the fact that it may take a little work to decide which library to use. That said, be careful not to let your choice of library or language guide your analysis!

Tip 2: Have a methodology ready

“If you can’t describe what you’re doing as a process, you don’t know what you’re doing.” – W. Edwards Deming

(*Translation Note 3) W. Edwards Deming (1900-1993) was an American statistician. After World War II, Deming spread quality control techniques based on statistical methods in Japanese industry, taking advantage of his involvement in the planning of Japan’s national census .

A methodology is essentially a framework that shapes the context in which research is guided . Simply restated in the context of an analytical project, a methodology keeps the project moving forward while you are working on it.

Having a methodology is important because it gives you a clear path to your goals . It’s also very useful when you have to explain your project to others.

A methodology is important because it gives you complete control over every step, from data acquisition to results presentation and other intermediate work.

A little-known benefit of having a methodology is that it makes you think more deeply about your project . For example, in a project analyzing racist violence in the US police force , I included the step ” Understand Your Bias “. Including this step allowed me to ensure that my findings were not subject to my own internal biases.

Yadhunath published in August 2020 the results of a project examining racist violence in US police using the Kaggle dataset Police Violence and Racial Equity. published. In this project, he took steps to identify his own racial biases. For him, who has never experienced racism, he learned about racism through movies and news. As a result of reflecting on his own knowledge of racism, he understood that some of it was wrong and why it was wrong.

Tip 3: Brainstorm like you used to in everyday life

“The best thing to do is get lots of good ideas and let the bad ones go.” – Linus Pauling

(*Translation Note 5) Linus Pauling (1901-1994) was an American biochemist. He is considered one of the pioneers of molecular biology. He is also known for his many famous quotes , and the quote above is one of them.

Early attempts to analyze a dataset can often lead to great confusion. Especially if you have a lot of instances and features, like the Kaggle Survey Challenge 2020 dataset .

Brainstorming is important in times of confusion. Brainstorming is, simply put, freeing your mind from physical (like papers) or digital (like computers) to come up with new ideas .

Although most definitions of brainstorming define it as a group process, there is research to support the theory that individual brainstorming produces better ideas than group sessions .

In the context of private brainstorming for a data analysis project, there are three steps that can help you get started.

  1. Read the dataset description – Think about what the main priorities are for the person who collected the data.
  2. Read Feature Descriptions – What are the best features in your view that fit the main priorities you considered in step 1?
  3. Read past work – If anyone has worked with the same or similar types of data in the past, review their work.

After you’ve done that, you’ll probably be able to write down all the ideas you come up with on a piece of paper (or digitally, if it’s not as old as my school). This kind of brainstorming is exactly what you do in everyday life. Write down ideas that you can use in data analysis before they disappear from your mind.

Note that brainstorming is the backbone of your analysis, so it should be descriptive.

Tip 4: Do preliminary analysis to identify the most promising narratives

“If you don’t think you know things, it’s because you know the first conditions and the first principles of what you want to know, and you don’t analyze it down to its simplest elements.” – Aristotle

After brainstorming, you may have multiple ideas that you want to pursue as narratives for your analysis. But if you want your work to be clear and impactful, you should choose just one main idea .

To make these choices, it is useful to write code to quickly bring up the system and perform preliminary analysis. Such work can be part of exploratory data analysis , and therefore we need to visualize the data at hand before focusing on the most promising stories to tell.

For example, in the 2020 Kaggle Survey Challenge on Machine Learning and Data Science, preliminary analysis showed that Indian respondents under the age of 21 had the fastest growth compared to other respondents in the survey . I discovered These preliminary analysis results helped me structure my project so that I could continue with the rest of the analysis.

(*Translation Note 6) Famous Australian blogger Tim Denning, in his Medium article, ” Why Focus on One Thing at a Time, ” lists the following four benefits of focusing on one goal: .

  1. It can reduce memory load .
  2. A single big goal is more meaningful to achieve than many small ones .
  3. If you focus on one goal, you also have one fear.
  4. Any conversation is easier to talk to because you can talk about it in the light of a single goal.

Tip 5: Use storyboards to build your narrative

“Storyboarding to me is a way of pre-visualizing an entire movie.” – Martin Scorsese

When it comes to storytelling with data, I personally consider Kohl Nussbaumer Knafrik to be one of the brightest minds in the field. The list of ideas she disseminates through her book and many of her talks is huge, but one that I find very helpful is the storyboarding process .

As professional writers say, a good story has 5 parts – they are the ‘Introduction’, ‘Rise’, ‘Climax’, ‘Descent’ and ‘Catastrophe’ . You can follow a similar structure to storyboard in the context of your data analysis project.

Start by introducing the data you have, then move on to the analysis or “why” statement of the main goal, analyze selected sub-goals as well, report your observations in a way that connects to the main goal, and finally all It combines these findings, selects the most important findings, and reports the findings as decisions that interested stakeholders can choose to make.

Cole Nussbaumer Knaflic is the founder of storytelling with data , a startup that helps people create data -driven stories, and a former Google employee. At Google, he was the manager of the data-driven recruitment unit, Peo Operation .

The five-part story referred to in the article is also called the “Freitag Pyramid ” because it was proposed by the German writer Gustav Freitag. The reason why it is called a pyramid is that the context of the story can be illustrated as a symmetrical triangle with the climax at the top (see also the image below).

Pyramid of Freitag

Image Source: Wikipedia ” Plot (Story) ” ” History ”

5 parts in story structure (Image Source: Author’s creation, inspired by this source )

Tip 6: It’s about them, not you

“It’s about them, not you.” – Clint Eastwood

From a more philosophical point of view, I encourage my readers to understand that when I analyze, the analysis is never just to satisfy my own eyes. The impact of analytical work is only related to how useful it is for the stakeholders involved.

Therefore, avoid inserting unnecessary graphs into your reports. There’s no need to show off that you’ve worked hard to insert it. Also, if the overall analysis isn’t coherent, you have to look in the trash can for discarded data and graphs.

It is also important to be able to link the outcomes of the analysis to actionable goals that stakeholders can choose.

Furthermore, don’t mislead your audience with meaningless visualizations .

VENNGAGE , a startup that develops and sells visualization tools, published a blog post in April 2020 titled ” 5 ways writers use misleading graphs to manipulate you. ” . The article describes five misleading ways to create charts, including:

American author Dan Rockwell , author of many books on leadership , wrote in his blog post , 7 Ways to Take Unfavorable Feedback: It preaches seven such attitudes.

  1. Any feedback is a gift that contributes to our success .
  2. Ask for feedback, not wait for it.
  3. Feedback you don’t like is usually helpful.
  4. Assuming that feedback that seems wrong is correct.
  5. Ask people who give you feedback, “Why is this feedback useful or important?”
  6. Even if you’re not asking for any suggestions , ask potential feedback people , “Do you have any suggestions?”

Tip 8: Pay attention to details

“The difference between good and great is attention to detail.” – Charles R. Swindall

(*Translation Note 12) Charles R. Swindall (1934-) is an American clergyman. He is known for running radio programs in 15 languages.

Even small things like the plot that develops the theme, the typography and colors you use in your report can have a big impact on how readers perceive your work.

A good example of the author’s attention to detail is Andrea Ort’s Birdcall Recognition EDA . This Kaggle project also has visualizations that match the bird’s color scheme.

Consistency is also a very important factor in analytical reporting. We don’t need flashy reports. You want something minimal that tells a story about the data analyzed.

(*Translation Note 13) In the Kaggle project ” Birdcall Recognition EDA ” published by Andrea Orto , a data scientist and Kaggle Notebook Master living in Romania, a large number of models were created to develop a model for identifying wild bird calls. Graphs are posted. The graph has an illustration of a beautiful wild bird (see image below). 

Other Helpful Tips

The links below contain tips from some of the best data analysts and storytellers who have had a huge positive impact on my data scientist journey so far.



Please enter your comment!
Please enter your name here

Recent Posts

Most Popular

Recent Comments