Dataset

This dataset was created by Daniil Krizhanovskyi

Released under MIT

Contents

Dataset

This dataset was created by Amy Okey

Released under RAIL (specified in description)

Contents

Dataset

This dataset was created by Under The Hood

Contents

Dataset

This dataset was created by Shubham Raj

Released under MIT

Contents

Dataset

This dataset was created by Saibhargav Ch

Released under Apache 2.0

Contents

Dataset

This dataset was created by Jhagdu

Contents

Dataset

This dataset was created by Jurgen

Contents

Dataset

This dataset was created by Mohsin Dahri

Released under CC0: Public Domain

Contents

Dataset

This dataset was created by Shubham Raj

Released under MIT

Contents

MLOps project for Arabic mental health text classification

This purpose of this (imaginary) study is to detect faulty satellite in order to prevent communication interruption. Understand how confirmation bias can lead to misleading results by utilizing a synthetic data set with a developed story line.

Problem Statement:

You are leading a team to conduct a study of the that will allow the space agency to predict the health status of satellites based on telemetry data to enable proactive maintenance and ensure optimal performance in space missions.

Stakeholders' Concerns:

1. Accurately predicting satellite health to minimize the risk of mission failures and optimize satellite usage.

2. Identifying the most critical factors that affect satellite health to focus on improving those aspects during the satellite design and maintenance process.

3. Reducing the rate of false positives and false negatives in predictions to avoid unnecessary maintenance efforts and ensure that actual issues are addressed promptly.

Misclassification Costs (estimation):

False Positive (predicting a malfunction when the component is healthy): Unnecessary maintenance check: $5,000 Unwarranted component replacement: $50,000

False Negative (predicting a component is healthy when it is malfunctioning): Data loss or degradation: $100,000 Partial mission failure: $500,000 Total mission failure or satellite loss: $300,000,000

Team Focus:

1. Thoroughly exploring the data to understand the relationships between various telemetry variables and satellite health.

2. Ensuring the model is accurate and reliable by selecting appropriate algorithms, performing feature engineering, and validating the model's performance using relevant metrics.

3. Identifying and addressing any data quality issues, such as missing values and incorrect data.

4. Investigating the importance of each variable in the prediction task and communicating these insights to stakeholders for better decision-making

DATA DICTIONARY:

Data Dictionary

1. time_since_launch (days)

Range: 0 to 3650 Description: Time since the satellite was launched.

1. orbital_altitude (km)

Range: 300 to 2000 Description: Altitude of the satellite's orbit.

1. battery_voltage (V)

Range: 20 to 30 Description: Satellite's battery voltage.

1. solar_panel_temperature (°C)

Range: -50 to 50 Description: Temperature of the satellite's solar panels.

1. attitude_control_error (degrees)

Range: 0 to 5 Description: Error in the satellite's attitude control system.

1. data_transmission_rate (Mbps)

Range: 10 to 100 Description: Rate of data transmission from the satellite to the ground station.

1. thermal_control_status (0 or 1)

Range: 0 (not working) or 1 (working) Description: Binary flag indicating if the thermal control system is working or not.

1. satellite_health (0 or 1)

Range: 0 (unhealthy) or 1 (healthy) Description: Target variable - binary flag indicating if the satellite is healthy or unhealthy

This dataset is a ground of whitepapers shared by Google in its AI workshop. It is a knowledgebase on various GenAI topics including prompt engineering, vector databases, embeddings, RAG, Agents, Agent companions, fine tuning and use of MLops in GenAI planning.

📄 Synthetic Resume Dataset (1,200 Records)

🧠 Overview

This dataset contains 1,200 high-quality, synthetic resume records carefully generated to simulate real-world professional profiles. It is ideal for HR analytics, recruitment model training, resume screening systems, and academic research.

The dataset is divided into:

• ✅ 60% IT Jobs (720 records)
• ✅ 40% Non-IT Jobs (480 records)

Each entry mimics realistic resume attributes across a wide range of roles—from fresh graduates to experienced professionals.

📊 Features

Each record includes the following 14 structured fields:

• Name, Age, Gender
• Education_Level, Field_of_Study, Degrees, Institute_Name, Graduation_Year
• Experience_Years, Current_Job_Title, Previous_Job_Titles
• Skills, Certifications, Target_Job_Description

👨‍💻 IT Job Titles (sample):

• Data Scientist, Cloud Engineer, Prompt Engineer, DevOps Engineer, Full-Stack Developer, MLOps Specialist, Quantum Computing Specialist, and more.

🧑‍⚕️ Non-IT Job Titles (sample):

• Nurse, Dentist, Financial Analyst, Mental Health Practitioner, Business Development Manager, Customer Success Manager, and others.

🔍 Key Highlights

• ✅ 100% synthetic data – no privacy concerns
• ✅ Zero missing values
• ✅ Realistic skill and certification pairings based on job roles
• ✅ Professionally written target job descriptions
• ✅ Balanced age (21–45) and experience (0–10 years) distribution
• ✅ Includes both freshers (36.8%) and working professionals (63.2%)

This Data is Extracted from coinGecko API with Price and Volume feature, data requires lot of feature engineering which is good practice for someone who wants to build MLOPs/Data Science project. This is static data set, if you want to make real pipeline this my repo where I have build real ETL pipeline and data stored in MongoDB Repo: https://github.com/Excergic/SOL-Price-Prediction Thank you.

Context

This data set contains combined on-court performance data for NBA players in the 2016-2017 season, alongside salary, Twitter engagement, and Wikipedia traffic data.

Further information can be found in a series of articles for IBM Developerworks: "Explore valuation and attendance using data science and machine learning" and "Exploring the individual NBA players".

A talk about this dataset has slides from March, 2018, Strata:

https://www.slideshare.net/noahgift/social-power-andinfluenceinthenba-89807740?qid=3f9f835a-f3d7-4174-8a8c-c97f9c82e614&v=&b=&from_search=1

Further reading on this dataset is in the book Pragmatic AI, in Chapter 6 or full book, Pragmatic AI: An introduction to Cloud-based Machine Learning and watch lesson 9 in Essential Machine Learning and AI with Python and Jupyter Notebook

Followup Items

• You can watch a breakdown of using cluster analysis on the Pragmatic AI YouTube channel

• Learn to deploy a Kaggle project into a production Machine Learning sklearn + flask + container by reading Python for Devops: Learn Ruthlessly Effective Automation, Chapter 14: MLOps and Machine learning engineering

• Use social media to predict a winning season with this notebook: https://github.com/noahgift/core-stats-datascience/blob/master/Lesson2_7_Trends_Supervized_Learning.ipynb

• Learn to use the cloud for data analysis.

Acknowledgement

Data sources include ESPN, Basketball-Reference, Twitter, Five-ThirtyEight, and Wikipedia. The source code for this dataset (in Python and R) can be found on GitHub. Links to more writing can be found at noahgift.com.

Inspiration

• Do NBA fans know more about who the best players are, or do owners?
• What is the true worth of the social media presence of athletes in the NBA?

Medical-Grade Explainable AI Project Assets

This dataset contains comprehensive assets for a production-ready Explainable AI (XAI) heart disease prediction system achieving 94.1% accuracy with full model transparency.

📊 CONTEXT: Healthcare AI faces a critical "black box" problem where models make predictions without explanations. This project demonstrates how to build trustworthy medical AI using SHAP and LIME for real-time explainability.

🎯 PROJECT GOAL: Create a clinically deployable AI system that not only predicts heart disease with high accuracy but also provides interpretable explanations for each prediction, enabling doctor-AI collaboration.

🚀 KEY FEATURES: - 94.1% prediction accuracy (XGBoost + Optuna) - Real-time SHAP & LIME explanations - FastAPI backend with medical validation - Gradio clinical dashboard - Full MLOps pipeline (MLflow tracking) - 4-Layer enterprise architecture

📁 ASSETS INCLUDED: - heart_clean.csv - Clinical dataset ready for analysis - SHAP summary plots for global explainability - Performance metrics and visualizations - Architecture diagrams - Model evaluation results

🔗 COMPANION RESOURCES: - Live Demo: https://huggingface.co/spaces/Ariyan-Pro/HeartDisease-Predictor - Notebook: https://www.kaggle.com/code/ariyannadeem/heart-disease-prediction-with-explainable-ai - Source Code: https://github.com/Ariyan-Pro/ExplainableAI-HeartDisease

Perfect for learning medical AI implementation, explainable AI techniques, and production deployment.