Data Science Portfolio Project - Churn prediction model

[@TopTimes_1]

Thanks! Your teaching Style, material and content is as beautiful as you are :) :P

[@karinadatascientist]

Thank you! 😃

[@badrinarayanans355]

Great Explanation

[@karinadatascientist]

Thank you

[@gaumeuvlog2603]

Thank you so much. I learn more when watching your videos 😊 Have a nice day ❤

[@karinadatascientist]

Aww, that's great to hear. Thank you

[@Tesseus2006]

I did notice you mentioned the meaning of the term tenure towards the end of your video meaning the length of time a costumer stayed with the company. Loved the video!

[@karinadatascientist]

Glad to hear that! I hope this video was useful

[@Tesseus2006]

@@karinadatascientist It really was. Thank you for making it and sharing your knowledge with us all.

[@jeffkamuthu3276]

A whole semester in half an hour! 🎉

[@karinadatascientist]

Aww. Great to hear. I am really trying to pick projects with real-world problems and data (not Titanic and Iris). I know I am not perfect and sometimes when review the video, think "I should have done this code differently", but I am also learning and practicing. I already learnt so much from this community: new libraries, new tools. Very grateful

[@jeffkamuthu3276]

Once the concept sinks in I think you can manipulate it with any data set.

[@aimadsadouk1304]

Again amazing new video

[@karinadatascientist]

Thank you for supporting me! I hope this video is useful

[@aimadsadouk1304]

It's very useful and magningul video

[@karinadatascientist]

@@aimadsadouk1304 glad to hear that

[@syedaafreen1075]

I appreciate your efforts and amazing concepts and thanks for this video

[@karinadatascientist]

Thank you for watching. It still amazes me that people decide to watch my video while they could have spend this time differently. Time is such a valuable resource we all have. I am very grateful for that.

[@JYFrimpong]

Thanks for this video. Very insightful and helpful! amazing

[@karinadatascientist]

Thank you. Glad to hear it is useful!

[@siddharth4873]

Thanks for accepting the request for this video. Keep up the good work!

[@karinadatascientist]

Thanks for the idea!

[@mapletech_22]

Awesome. Thanks 👍 🎉

[@karinadatascientist]

Thank you for watching!

[@furkanyigt1147]

Hello Karina, you're amazing! Could you please create a comprehensive and distinctive project for data scientists to help them land their first job? We really need this.

[@karinadatascientist]

thank you. Ok, let me find good dataset and create another project. However, churn model, logistic regression and k-means that I did before are good for your resume, you just need to find a more complicated data set, create good README file and you can add it to your GitHub

[@hoangha6680]

thanks for your video. Hope that your future videos will also include model deployment as well ^^

[@karinadatascientist]

At some point I might add deployment

[@ibrahimdenisfofanah6420]

You’re just the best. I would appreciate it if you could mentor me

[@karinadatascientist]

Thank you. I do paid 1:1 coaching, if you are interested

[@TopTimes_1]

@@karinadatascientist This is business :)

[@alexmendes-89]

Very good😁

[@karinadatascientist]

Thank you

[@magbomajoshua1585]

Great content, please can you do a video on pipeline in machine learning

[@karinadatascientist]

Thank you, unfortunately that more of a data engineering / ML engineering territory. I never worked on pipelines

[@magbomajoshua1585]

@@karinadatascientist okay, but I see it machine learning solutions.

[@emsif]

this was very helpful, thanks! But i´m everytime confused because i´ve never seen a video so far in which the assumptions of linearity between the features and the log odds of the target or multicollinearity in a logistic regression model were checked. Is there´s a reason? Or why didn´t you check these assumptions?

[@karinadatascientist]

It is a vild point. Have you worked as a data scientist in the business or you worked in academia/as a researcher? It’s quite common for people not to check these assumptions in practical applications of logistic regression. In practical data science tasks, the focus is often on predictive power, and models like decision trees, random forests, or gradient boosting might be preferred, where assumptions of linearity are not required. Logistic regression will still run even with multicollinearity, and in many real-world datasets, multicollinearity doesn’t prevent the model from being useful, particularly when the goal is prediction rather than inference. Since in real-life projects main focus is accuracy - I haven't run into issues where these assumptions caused major problems.

[@giraffa-analytics]

:D sometimes you give up and cry! That is so real.

[@karinadatascientist]

hahah. I wondered if anyone would hear this part or people just skip it. Sometimes nothing helps: you do feature engineering, parameter tuning, generate more data, clean it, and accuracy is 52%. Had it with my recommendation engine, very unbalanced data and I couldn't tame it

[@giraffa-analytics]

@@karinadatascientist right... I've been a data scientist for 10 years and I've come to the conclusion that resilience is the number 1 skill I master. 😅😂

[@karinadatascientist]

@@giraffa-analytics That's a good one. I will use this phrase!

[@anuvabkumarrit4669]

❤🥺

[@karinadatascientist]

Thank you

[@anuvabkumarrit4669]

@@karinadatascientist I was wondering when you will open a KZbin channel. Very happy to learn new things 🤭

[@karinadatascientist]

@@anuvabkumarrit4669 :) I've been talking about KZbin for a year, even spoke to a YT producer, who said it will cost me $4 000 USD per episode (if we film with 2-3 cameras, lights, with proper studio, professional edits etc). I just couldn't justify such spend for something that is a passion project. Then people on TikTok started encouraging me to start, and there were days when people asked me daily if I have a channel. So I decided to give it a try. I film on a phone, so I know the quality is not there, I edit myself (hopefully that skill is improving a bit), but I hope that these videos will find its audience anyway

[@anuvabkumarrit4669]

@@karinadatascientist I think the quality is good only so far, in fact if you haven't mentioned you did this through phone, no one would guess. In the end, it's not just the quality of video but the content that is shared. I have been following your Instagram even you cleared a lot of doubts for me over direct message. So I am very content :)

[@PreplyshortVid]

Slow Down OMG Well done Karina. Please my feedback is Slow down. Take slow. We love you already. I love you so much so please Slow down so I can assimilate better. Haaaaaaaaaaa

[@karinadatascientist]

hahahaha. Amazing comment, it made my day. Thank you.