codaks

And incase you don't know, I'm learning ML engineering, I trade memecoins(out of liquidity), I write technical threads on projects I feel might be an issue to non techy testers, and other projects i find interesting and I'm a web3 tester too. Onchain Based🔹️

I built a simple tax notification automated workflow A form connecting to Google Sheets to store candidates' data and a switch between to pick if the candidate is to pay tax or not but sending them an email.

@eniola_yun27360 That's 95% to 5% Explain why, after using SMOTE, you still have a report like that, upsampling has a high chance of noise, downsampling = losing valuable data. Try XGBoost, RandomForest classifier, and Balanced Bagging classier and Balanced RandomForest classifier

@Codaksmade 19:1 I'll. Thanks

I tuned this model but decrease in performance instead. What could be wrong?

@eniola_yun27360 What's the ratio of the imbalance You can also try XGBoost without applying SMOTE

@Codaksmade I used it before training, what other suggested techniques

@DJiafei I would say it's an immediate sibling with Mechatronics Engineering

People often ask whether robotics belongs to mechanical engineering, electrical engineering, or computer science. Honestly, it is all of them. Robotics is the ultimate interdisciplinary field, and that is what makes it so powerful—from agriculture to protein discovery in wet labs.

Wrote**

Write another thing, a Continuation: @Codaks/why-this-model-4cae17763f18" target="_blank" rel="nofollow noopener">medium.com/@Codaks/why-th…

@RoseMarvelous4 No pressure abeg

We shall not!!!

The funniest thing is, I honestly don't know what I'm doing

@ChinenyeMba2 Very much me....still trusting the process and keeping a clear mind It was only a thin line between me selling my PC and not selling a few days ago. Life happens fr, just keep trusting God Use the saying: "Send the bill to ABBA, he go run am Nah you dey fear cos of the amount"

This is a testament that situations change when there is a focused plan. About 8 weeks ago, no one could see this space and believed there was a plan ongoing. It was devastating to even look at. If you know me, you know I don’t function well in disorganized spaces. I love my environment neat, clean, and put together. My previous apartments? Always comfortable for me. So you can imagine what it felt like managing this space for weeks. It was driving me nuts. I even saw a post on Facebook where someone said, “Arrange your clothes first.” 😂 In my mind, I just laughed… “If only you knew.” You just can’t be as organized as me. But here’s what people didn’t know… I had already seen the end. That’s why I made a post with this same space earlier. I needed a Before, because I knew I would come back with an After. When the vision is clear, you are already halfway certain of the outcome — even when your current reality looks nothing like it. No one saw the work I was putting in. No one saw the patience it required. No one saw the time, the money, the discomfort. I had to explain my vision to my cousin and my baby sis, and they were like, “This is too much.” 😭 But only I understood what it would take. Not them. Not you. So I stayed with it. Focused. Patient. Intentional. And now? The same thing that looked like “nothing” is what people are applauding today. Funny how that works. The truth is, things can change — sometimes faster than people expect. So don’t let the pressure of “visible results” rush you into something you’re not ready for. Don’t be ashamed of your process. You will need that proof. And please… don’t be visionless. Stop managing life. Sometimes, you step out of that “managing” room and build what you actually want to see. As far as your eyes can see. #Congratulations to me once again. I couldn’t be prouder. What did you learn from my story? Tell me 🌚 Before After

@timmyhasagoatee O ju ri

omo💔

@dlifeof_jay Thank you, Senor.

@Codaksmade I just saw your formulae again, a rule based won’t work on inference because of the included actual values Yes bro, I feel you should go ahead with this technique

Nice work bro! I love the idea of thresholding the regression outputs. for the classification if possible, I suggest you make the outputs binary instead of a 3 class output, that can help with the imbalance or for now instead of a model you can define your current 3 way output as a rule system Because if I suggest sourcing more data now, based on your type of project the probability of the dataset still being imbalanced is high.

@dlifeof_jay, what would you have suggested? Trust me, I have tried various techniques (random forest, xgboost, baggingclassifer, balanced randomforest, easy ensemble classifer), even tempered with the features, but still got imbalanced datasets which is the main cause of this

having picked a suitable regression model to get the PR, the next issue I have been facing is the classification model The data suffers from extreme imbalance,explains the reason why sampling isn't an option, it's either the recall is high and precision is low or vice versa.

Actually, i'm learning a lot of WHY on this project

Started a new thing.....expect more in the coming days This is a blog post automation process, The post is made by a condition clause on the spreadsheet,the spreadsheet takes in the blog idea, the context and the conditional clause to make the post.

just on this project, I have got to learn about more models and understand why metrics are important in decision making

Unarguably, the question is not which The question is WHY why use this technique instead of another?

In the name of the Father, and of the Son and of the Holy Spirit Amen.

@nicoletteduclar @tempaitown 327939

more on @tempaitown. it's a massive multi-agent feudal sim game. fans of CIV/AoE? gotchu agents become lord sovereign by owning lands, trading goods, fighting rogue agents & recruiting subagents to work their plots first lands will be dropped to 9k Farcaster ID reply your FID or make sure to have X connected via Warpcast Tempo has native gas sponsorship so you don't even need to bridge. UX will be amazing. on how to mint soon