I was training credit card fraud detection model with random forest .
The above image is confusion matrix which is evaluation metrics and can see the model has good performance over test data with small false positive ( classifying fraud data as non ). The data I used is cleaned data you can practice hyper parameter tuning. https://www.kaggle.com/datasets/nelgiriyewithana/credit-card-fraud-detection-dataset-2023

ow I forgot to tell in python it is only compatible with python11 and below

By the way, I haven’t posted anything this week because I was catching up on classwork—I was really behind. I've also been becoming more socially active, talking to my classmates and making new friends. Honestly, I like that part. But one thing I regret is messing up my sleep schedule, for real. Aside from that, I’ve only made small progress. What I want to say is that it’s important not to fall behind, catching up is really costly. Good night🚶‍♀‍➡️

https://youtu.be/_LbeAYqCnH4?si=aPCPJ3zUR-Au-KiT I went to the same high school. I am proud that he shines in the game industry.🔥

Check out the xverse library for your machine learning project, it helps you select and prepare the most effective features for modeling.

MLflow is a must-have tool. It helps track experiments, log metrics, and manage model versions all in one place. no need to re-run notebooks 🤌

I was wondering about vs code and I found this article if u r curious... here 🤌 https://www.linkedin.com/pulse/from-origin-optimization-story-visual-studio-code-thirumoorthy-bqtxc/?trackingId=UWxtIt85Te2Q4g4fjrjrog%3D%3D it is short tho

Today, I came across RAG (Retrieval-Augmented Generation) while working on an assignment for 10 Academy.
You know, Large Language Models (LLMs) like GPT or DeepSeek are incredibly powerful, but they do have limitations—especially when it comes to domain-specific expertise or producing accurate responses in specialized contexts.

That’s where RAG comes in. It enhances LLMs by retrieving preprocessed documents from a specific domain (like company data or product manuals), and then integrates this information into the generation process. For example, if you're building a customer service chatbot, RAG can retrieve relevant company documents and use them to generate more accurate, context-aware answers.

Just completed Week 1😊 ... still in development

Spam 😬, you may not even listen to music, but the lyrics in this song are about life. I’ve been listening to it on repeat , it really hits hard. Check it out https://youtu.be/RPpoZYt1QME?si=HzVNItYoxS1W22sV"

📣 Excited to Announce!

WTM Addis and GDG Addis are teaming up with NSK.AI for an incredible AI Bootcamp:

“Intro to AI Agents: From RAG to Deployment”

This FREE, beginner-friendly Bootcamp will help students, developers, and AI enthusiasts build real-world AI apps that can answer questions using their own documents or external knowledge. All hands-on with Langchain and other open-source AI tools!

As a participant, you’ll gain:

✅ Hands-on RAG (Retrieval-Augmented Generation) experience
✅ Practical projects with Langchain & open-source tools
✅ Text & audio AI interactions
✅ A portfolio-ready AI project
✅ Support from a vibrant learning community
✅ Knowledge of vector databases, retrieval strategies, prompt engineering & more

No advanced AI background needed; just curiosity and basic Python skills!

🗓 Starts: 19th July 2025 | 💻 Self-paced | 💰 FREE

Register here 👉 https://forms.gle/1uM8go8yAPpr46KLA to start building real AI apps!

So, week 2 I didn’t touch the IDE project. I started the Play Store app review project instead. It been almost 2 months since I joined the 10Academy AI Kifiya Mastery course. The thing is it is project based learning, and my week 2 project was to do Play Store review analysis for three selected mobile banking apps. Then, in week 6, the project was using RAG to build a customer complaint chatbot. So the idea for this project came to me like by combining week 2 and week 6 project why not owners get actionable insights from users review just by using play Store app id

This is also in progress ... for improvements and my backend is like 👨‍🦽

Starting tomorrow I will post what I gained through this training.

good night ☺️

Always you have been told that work is a curse and labour a misfortune.
But I say to you that when you work you fulfill a part of earth's furthest dream,
assigned to you when that dream was born,
And in keeping yourself with labour you are in truth loving life,
And to love life through labour is to be intimate with life's inmost secret.

from the

The Prophet book

-- Khalil Gibran

Z-score
This is standard normal distribution. The 0 point is the mean of the data and it has standard deviation 1. A data above the average has positive Z-Score and negative for below.

Z-Score is calculated like this

Z = data point - mean of the data
/ Standard deviation

So this indicate how far our data point is from the mean.

The probability of a z-score tells you how likely a value is to appear in a normal distribution and from the image you can see zscore that fall between 3 and -3 have 99.9 probability to be in normal distribution.

Here the use of zscore comes in
Outlier is a data point that is unusual. like we have data of high school students and it is like getting 80 or 5 in age column and the other data is clustered between 15-20.
Outliers affect your model performance.
So the data point with z-score above absolute value of 3 or -3 has 0.1 % chance to be found in normal distribution of the data. So in the another word it is outlier.

#SharingisCaring 😊