XGBoost is now my go-to number 1 must have algorithm in my data science toolkit. But for years, I had no clue what I was doing. In 3 minutes, I’ll share 3 months of research (business case included). Let’s go: 1. XGBoost, which stands for Extreme Gradient Boosting, is an advanced implementation of the gradient boosting machine (GBM) algorithm. It was developed to optimize both computational speed and model performance. 2. Gradient Boosting Machine (GBM): GBMs are an ensemble approach that combines multiple weak learners (typically decision trees) to create a strong predictive model. 3. Difference between GBM and Random Forest: Random Forest also uses decision trees. However, the difference is how the trees are developed. GBM sequentially adds weak learners (shorter trees), where each one corrects its predecessor. RF constructs strong learners (large trees) in parallel using Bootstrap Aggregation (Bagging). 4. Performance: XGBoost is an ultra-fast implementation of GBM that includes high-efficiency, scalability, ability to handle sparse data, parallel learning, and regularization to reduce over-fitting. XGBoost tends to be more efficient than Random Forest and traditional GBM, and often provides better performance. This is why I like it so much. 5. Business Use Cases: I use XGBoost in many business cases. Let’s start with the one that made the most impact, a $12,000,000 sales increase. Lead scoring. In 2 years of developing a Lead Scoring Model that made my company $12,000,000, I used a number of different algorithms. I started with the most basic- Logistic Regression for classification probability of a customer purchasing. And over time, improved it with better algorithms, better and more complex features. The final iteration had XGBoost as a key model in the algorithm. 6. Business Use Case (Time Series): I later discovered that I could use XGBoost on time series for forecasting sales demand at the product level. This was a major improvement over less-scalable techniques (ARIMA, ETS) that had to be run iteratively on every product. We had 12,000+ products. XGBoost cut training times from 3 days to about 4 minutes. One thing to watch out for is that because it’s tree-based, XGBoost cannot predict beyond the maximum or below the minimum in a dataset. So differencing may be required. === Need help improving your data science skills? I'd like to help. Here's how: 👉 My Free 10 Skills Webinar: I put together a free on-demand workshop that covers the 10 skills that helped me make the transition to Data Scientist: learn.business-science.io/fr… 👉 Learn ChatGPT for 10X Faster DS Projects: I have a live workshop where I'll share how to use ChatGPT for Data Science (so you can complete projects 10X faster): learn.business-science.io/re… If you like this post, please reshare ♻️ it so others can get value (follow me, 🔥 Matt Dancho 🔥 for more data science concepts).

I had a wild 6-month journey in developing my first deep learning model in 2015. Here's 6 months of research on Neural Networks in 6 minutes. Let's go. 1. What is a Neural Network (the Neuron): A computational model inspired by the way neural networks in the human brain work. It's structured like a brain with neurons. Neurons get activated via weights and activation functions. 2. Layers: There are three main types of layers: input layer, hidden layers, and output layer. The input layer receives the input data, hidden layers process the data, and the output layer produces the final result. 3. Weights and Biases: Connections between neurons are associated with weights and biases. A weight determines the strength of the connection between two neurons. During the training process, the network adjusts these weights and biases based on the error of the output compared to the desired outcome. 4. Activation Functions: Neurons apply activation functions to the data they receive. These functions are crucial for the neural network's ability to make complex patterns or data visible to the learning algorithm. Common activation functions include sigmoid, tanh, and ReLU (Rectified Linear Unit). 5. Learning: In the training phase, the network learns by adjusting the weights and biases. The most common learning method is backpropagation combined with an optimization algorithm like gradient descent. The network makes predictions, compares them to the actual outcome, and adjusts the weights and biases to minimize errors. 6. Types of Neural Networks: There are 10 that I've experimented with (I'll explore these in a separate post). Common are CNN, RNN. 7. Business Use Cases: Deep Learning has been most important in Time Series, which is the number 1 area I've gotten decent results. I've wasted A LOT OF TIME (6-months) trying to get Deep Learning models to work well on Tabular Data (and it couldn't beat a Logistic Regression I built in an hour). Usually, XGBoost crushes Deep Learning on tabular data (it's both faster and better suited). Non-tabular data (unstructured text, image data), Deep Learning tends to be the way to go. That's a wrap on Deep Learning for today. Make sure to review Point 7 on business use cases. I once wasted 6 months on a DL model that performed worse than a Logistic Regression that I built in a day. === There’s a lot more to learning Data Science for Business. I’d like to help. I put together a free on-demand workshop that covers the 10 skills that helped me make the transition to Data Scientist: learn.business-science.io/fr… And if you'd like to speed it up, I have a live workshop where I'll share how to use ChatGPT for Data Science: learn.business-science.io/re… If you like this post, please reshare ♻️ it so others can get value.