{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"
\n π Machine Learning Grandmaster\n
","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19"}},{"cell_type":"markdown","source":"\n Welcome to ML Grandmaster, where we'll take a look at the most important Machine Learning algorithms and fundamental concepts from the ground up.
\n We'll look at the theory behind the concepts and also the code to implement these ideas in the real world.\n
","metadata":{}},{"cell_type":"markdown","source":"## ML Grandmaster Notebooks\n
\n","metadata":{}},{"cell_type":"markdown","source":"\n π³ Gradient Boosting (XGBoost)\n
","metadata":{}},{"cell_type":"markdown","source":"\n In this ML Grandmaster notebook, we'll be taking a look at Gradient Boosting from the ground up, and overviewing XGBoost (eXtreme Gradient Boost), one of the most popular Gradient Boosting algorithms.
\n
","metadata":{}},{"cell_type":"markdown","source":"![extreme](https://media2.giphy.com/media/uaOqWaZpebl8k/giphy.gif?cid=790b76112d4331bddab4d09906a33b08afb80497911f4b8c&rid=giphy.gif&ct=g)","metadata":{}},{"cell_type":"markdown","source":"# π Index\n* [π Prerequisites and Notes](#0)\n* [π€¨ Intuition](#1)\n* [π Import](#2)\n - [Packages](#2.1)\n - [Dataset](#2.2)\n* [π Gradient Boosting Algorithm](#3)\n* [π Loss Function](#4)\n* [𧱠Building the Boosted Tree](#5)\n - [Step 1](#5.1)\n - [Step 2A](#5.2)\n - [Step 2B](#5.3)\n - [Step 2C](#5.4)\n - [Step 3](#5.5)\n* [π₯ XGBoost Overview](#6)\n* [π» Code Implementation](#7)\n* [π Conclusion](#8)","metadata":{}},{"cell_type":"markdown","source":"\n# π Prerequisites and Notes","metadata":{}},{"cell_type":"markdown","source":"","metadata":{}},{"cell_type":"markdown","source":"\n Gradient Boosted Trees are some of the most popular Ensemble Learning models, and some of the best models, performance wise, on large tabular datasets.
\n XGBoost is one of the most popular Gradient Boosting algorithms, as we can see from the winning kernels of a lot of Kaggle competitions!\n
","metadata":{}},{"cell_type":"markdown","source":"\n Learning how Gradient Boosting works, and learning about one of its most popular implementations XGBoost, will give us some insights into real world predictive models in High Energy Physics, Recommendation Systems, Ad Click Through Rate Predictors and so on.\n
","metadata":{}},{"cell_type":"markdown","source":"\n# π€¨ Intuition","metadata":{}},{"cell_type":"markdown","source":"\n The fundamental idea behind Gradient Boosted Trees is that it is an Ensemble Learning technique (so $ N $ trees are used) where each tree tries to improve the result of the previous tree, by optimizing the difference between its predictions and the output value.
\n This differs from
π³ Random Forests, which is still an Ensemble Learning method, but there, each tree makes a prediction independently from the other trees, and the final prediction is an aggregation of the predictions of each tree.
\n The image below gives us an idea of how a prediction is made by a Gradient Boosted Tree.\n
","metadata":{}},{"cell_type":"markdown","source":"![1.png](attachment:404346b2-e5a1-4f1b-b576-9b32cfd638d9.png)","metadata":{},"attachments":{"404346b2-e5a1-4f1b-b576-9b32cfd638d9.png":{"image/png":"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"}}},{"cell_type":"markdown","source":"\n Note: these individual trees are usually called Base Estimators or Weak Estimators.
\n We'll walk through the Gradient Boosting Algorithm in this notebook for a Regression problem, and see how it works on a simplified version of the House Prices dataset.\n
","metadata":{}},{"cell_type":"markdown","source":"\n# π Import","metadata":{}},{"cell_type":"markdown","source":"\n## Packages","metadata":{}},{"cell_type":"code","source":"import numpy as np # linear algebra\nimport pandas as pd # data manipulation\n\nfrom xgboost import XGBRegressor # model\n\nfrom sklearn.tree import DecisionTreeRegressor # model (used in Building Tree walkthrough)\nfrom matplotlib import pyplot as plt # visualization\nfrom sklearn.tree import plot_tree # tree visualization","metadata":{"execution":{"iopub.status.busy":"2022-10-09T20:00:17.800313Z","iopub.execute_input":"2022-10-09T20:00:17.80078Z","iopub.status.idle":"2022-10-09T20:00:18.706664Z","shell.execute_reply.started":"2022-10-09T20:00:17.800692Z","shell.execute_reply":"2022-10-09T20:00:18.705373Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"np.random.seed(42) # set random number generator seed (for reproducibility)","metadata":{"execution":{"iopub.status.busy":"2022-10-09T20:00:46.937928Z","iopub.execute_input":"2022-10-09T20:00:46.938353Z","iopub.status.idle":"2022-10-09T20:00:46.944943Z","shell.execute_reply.started":"2022-10-09T20:00:46.938317Z","shell.execute_reply":"2022-10-09T20:00:46.943746Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n## Dataset","metadata":{}},{"cell_type":"code","source":"df = pd.read_csv('../input/house-prices-advanced-regression-techniques/train.csv')","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:24.696456Z","iopub.execute_input":"2022-10-07T15:29:24.696834Z","iopub.status.idle":"2022-10-07T15:29:24.751081Z","shell.execute_reply.started":"2022-10-07T15:29:24.696792Z","shell.execute_reply":"2022-10-07T15:29:24.749629Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def preprocess(df):\n cols_to_drop = [col for col in df.columns if col not in ['LotArea', 'SalePrice']]\n df['Pool'] = (df['PoolArea'] > 0).astype('int64')\n df['NewConstruction'] = (df['SaleType'] == 'New').astype('int64')\n df['SalePrice'] = df['SalePrice'] // 1000\n df = df.drop(columns=cols_to_drop)\n df = df[['NewConstruction', 'Pool', 'LotArea', 'SalePrice']]\n \n return df","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:24.755803Z","iopub.execute_input":"2022-10-07T15:29:24.756384Z","iopub.status.idle":"2022-10-07T15:29:24.762929Z","shell.execute_reply.started":"2022-10-07T15:29:24.756347Z","shell.execute_reply":"2022-10-07T15:29:24.761681Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df = df.iloc[[197, 810, 11, 18, 1449], :]\ndf = preprocess(df)\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:24.764354Z","iopub.execute_input":"2022-10-07T15:29:24.764688Z","iopub.status.idle":"2022-10-07T15:29:24.803114Z","shell.execute_reply.started":"2022-10-07T15:29:24.764658Z","shell.execute_reply":"2022-10-07T15:29:24.801987Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n# π Gradient Boosting Algorithm","metadata":{}},{"cell_type":"markdown","source":"\n The Gradient Boosting Algorithm is what allows us to build a Gradient Boosted Tree. Each Gradient Boosted Tree implementation (i.e. XGBoost or LightGBM) uses different versions of this algorithm, but they all follow this general guideline.
\n Let's define the algorithm below.\n
","metadata":{}},{"cell_type":"markdown","source":"\n Given an input variable $ x_i $, an output variable $ y_i $ where $ i \\in \\{1, ... , n\\} $ where $ n $ is the number of samples.
\n Given a differentiable
Loss Function $ L(y,F(x)) $.
\n Given a
Number of Base Estimators $ M $.
\n Given a
Learning Rate $ \\nu $ so that $ 0 < \\nu \\leq 1 $.
\n
\n - Initialize the model with a constant value $F_0(x) = \\underset{\\gamma}{\\arg\\min} \\sum_{i=1}^n L(y_i, \\gamma)$ (for Regression, this will usually be the mean of output values)
\n - \n For $ m = 1 $ to $ M $\n
\n - \n Compute the so-called pseudo-residuals:\n $$ r_{im} = -\\left[\\frac{\\partial L(y_i, F(x_i))}{\\partial F(x_i)}\\right]_{F(x)=F_{m-1}(x)} \\quad \\mbox{for } i=1,\\ldots,n $$\n
\n - \n Fit a CART (π³ Classifier or π³ Regressor) Base Estimator (or Weak Estimator) to pseudo-residuals, trained using the training set $ \\{(x_i, r_{im})\\}_{i=1}^n $, to create residual estimates $ \\gamma_{im} $.\n
\n - \n Update the model: \n $$ F_{m}(x)=F_{{m-1}}(x) + \\nu \\cdot \\gamma_m $$\n
\n
\n \n - \n $ F_M(x) $ will be our final predictor.\n
\n
\n
","metadata":{}},{"cell_type":"markdown","source":"![huh](https://i.imgflip.com/1myuho.jpg)","metadata":{}},{"cell_type":"markdown","source":"\n Let's clear this math up by going through the algorithm step by step.\n
","metadata":{}},{"cell_type":"markdown","source":"\n# π Loss Function","metadata":{}},{"cell_type":"markdown","source":"\n The Loss function is the heart of our algorithm. In Step 2A, we take the Gradient of the Loss function to optimize our Tree.
\n This is why it's called Gradient boosting.\n
","metadata":{}},{"cell_type":"markdown","source":"\n Given an observed value $ y_i $ and an estimated value $ F(x_i) $ let's take our Loss Function to be:\n $$ L(y_i, F(x_i)) = \\frac{1}{2} (y_i - F(x_i))^2 $$\n
","metadata":{}},{"cell_type":"markdown","source":"\n This Loss Function allows us to compute the pseudo-residuals very easily, since when we take its derivative:\n $$ r_{i} = -\\left[\\frac{\\partial L(y_i, F(x_i))}{\\partial F(x_i)}\\right] = -\\frac{\\partial}{\\partial F(x_i)} \\frac{1}{2} (y_i - F(x_i))^2 = $$\n $$ = - (-1) \\frac{2}{2}(y_i - F(x_i)) = y_i - F(x_i)$$\n The pseudo-residuals end up being standard residuals, a.k.a the difference between observed value and estimated value. Very easy to compute.
\n There are other Loss Functions we could choose, but we'll go with this one.\n
","metadata":{}},{"cell_type":"markdown","source":"\n# 𧱠Building the Boosted Tree","metadata":{}},{"cell_type":"markdown","source":"\n Let's go through the Gradient Boosting Algorithm and build our Gradient Boosted Tree, using our example dataset.\n
","metadata":{}},{"cell_type":"code","source":"df","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:34.472081Z","iopub.execute_input":"2022-10-07T15:29:34.472603Z","iopub.status.idle":"2022-10-07T15:29:34.485003Z","shell.execute_reply.started":"2022-10-07T15:29:34.47256Z","shell.execute_reply":"2022-10-07T15:29:34.483834Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n In our example dataset, the y variable (what we want to predict) will be SalePrice, while the x variables will be NewConstruction, Pool, LotArea.\n
","metadata":{}},{"cell_type":"markdown","source":"\n## Step 1","metadata":{}},{"cell_type":"markdown","source":"\n Step 1: Initialize the model with a constant value $F_0(x) = \\underset{\\gamma}{\\arg\\min} \\sum_{i=1}^n L(y_i, \\gamma)$ (for Regression, this will usually be the mean of output values).
\n So let's take the mean of SalePrice.\n
","metadata":{}},{"cell_type":"code","source":"df['F0'] = df['SalePrice'].mean()\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:35.410231Z","iopub.execute_input":"2022-10-07T15:29:35.411298Z","iopub.status.idle":"2022-10-07T15:29:35.426226Z","shell.execute_reply.started":"2022-10-07T15:29:35.411254Z","shell.execute_reply":"2022-10-07T15:29:35.424663Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n Now we start with Step 2. To keep things simple, let's keep set the Number of Base Estimators to $ M = 2 $.
\n So we'll iterate 2 times, $ m $ will go from 1 to 2.
\n Let's go through iteration 1 in the next sections.\n
","metadata":{}},{"cell_type":"markdown","source":"\n## Step 2A","metadata":{}},{"cell_type":"markdown","source":"\n Step 2A: Compute the so-called pseudo-residuals:\n $$ r_{im} = -\\left[\\frac{\\partial L(y_i, F(x_i))}{\\partial F(x_i)}\\right]_{F(x)=F_{m-1}(x)} \\quad \\mbox{for } i=1,\\ldots,n $$\n
","metadata":{}},{"cell_type":"markdown","source":"\n We start our first iteration $ m = 1 $. As we've calculated in the Loss Function section, our pseudo-residuals $ r_{i1} $ turn out to be very simply standard residuals:\n $$ r_{i1} = y_i - F_0(x_i) $$\n
","metadata":{}},{"cell_type":"code","source":"df","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:29:38.963826Z","iopub.execute_input":"2022-10-07T15:29:38.964338Z","iopub.status.idle":"2022-10-07T15:29:38.977761Z","shell.execute_reply.started":"2022-10-07T15:29:38.964297Z","shell.execute_reply":"2022-10-07T15:29:38.976332Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n In our example dataset $ y_i $ are the samples in the SalePrice column, and $ F_0(x_i) $ is the column F0 we calculated in Step 1.
\n So now let's calculate the residuals, which we'll call r1 in the dataframe.\n
","metadata":{}},{"cell_type":"code","source":"df['r1'] = df['SalePrice'] - df['F0']\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T15:55:04.232752Z","iopub.execute_input":"2022-10-07T15:55:04.23322Z","iopub.status.idle":"2022-10-07T15:55:04.248222Z","shell.execute_reply.started":"2022-10-07T15:55:04.233175Z","shell.execute_reply":"2022-10-07T15:55:04.246774Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n## Step 2B","metadata":{}},{"cell_type":"markdown","source":"\n Step 2B: Fit a CART (
π³ Classifier or
π³ Regressor) Base Estimator (or Weak Estimator) to pseudo-residuals, trained using the training set $ \\{(x_i, r_{im})\\}_{i=1}^n $, to create residual estimates $ \\gamma_{im} $.\n
","metadata":{}},{"cell_type":"markdown","source":"\n In this step we fit a
π³ Decision Tree Regressor (if this were a Classification problem, we'd use a
π³ Decision Tree Classifier).
\n We won't go into much detail in this step, since we have 2 whole notebooks in the π ML Grandmaster series dedicated to these topics.
\n For this step we'll leverage sklearn's DecisionTreeRegressor to make the predictions $ \\gamma_{i1} $.\n
","metadata":{}},{"cell_type":"code","source":"baselearner1 = DecisionTreeRegressor(\n criterion='squared_error',\n splitter='best',\n min_samples_split=4\n)","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:07:54.698593Z","iopub.execute_input":"2022-10-07T16:07:54.699144Z","iopub.status.idle":"2022-10-07T16:07:54.705151Z","shell.execute_reply.started":"2022-10-07T16:07:54.699098Z","shell.execute_reply":"2022-10-07T16:07:54.703803Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"baselearner1.fit(\n X=df[['NewConstruction', 'Pool', 'LotArea']],\n y=df['r1']\n)","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:08:55.475947Z","iopub.execute_input":"2022-10-07T16:08:55.476471Z","iopub.status.idle":"2022-10-07T16:08:55.493126Z","shell.execute_reply.started":"2022-10-07T16:08:55.47643Z","shell.execute_reply":"2022-10-07T16:08:55.491766Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df['gamma1'] = baselearner1.predict(X=df[['NewConstruction', 'Pool', 'LotArea']])\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:11:33.493535Z","iopub.execute_input":"2022-10-07T16:11:33.494389Z","iopub.status.idle":"2022-10-07T16:11:33.51329Z","shell.execute_reply.started":"2022-10-07T16:11:33.494342Z","shell.execute_reply":"2022-10-07T16:11:33.512228Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n## Step 2C","metadata":{}},{"cell_type":"markdown","source":"\n Step 2C: Update the model: \n $$ F_{m}(x) = F_{{m-1}}(x) + \\nu \\cdot \\gamma_m $$\n
","metadata":{}},{"cell_type":"markdown","source":"\n For this step, we need to talk about the Learning Rate $ \\nu $ (pronounced \"nu\". However the XGBoost documentation refers to it as \"eta\" which is the pronunciation of the greek letter $ \\eta $. I don't know why they did this).
\n The Learning Rate is a number greater than 0 but smaller than 1, and it must be thought of as parameter that modulates by how much each Gradient Boost step must move in the right direction.
\n The closer to 1, the bigger the step. The closer to 0, the smaller the step.\n
","metadata":{}},{"cell_type":"markdown","source":"\n For our example, let's choose $ \\nu = 0.1 $.
\n Note, it has been empirically shown that a small Learning Rate $ \\nu $ and a large number of Base Estimators $ M $ yields better performance than a large Learning rate and a small number of Base Estimators $ M $. This however comes at a computational cost, since computational complexity scales with $ M $.\n
","metadata":{}},{"cell_type":"code","source":"nu = 0.1\n\ndf['F1'] = df['F0'] + nu * df['gamma1']\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:25:55.88855Z","iopub.execute_input":"2022-10-07T16:25:55.888979Z","iopub.status.idle":"2022-10-07T16:25:55.907273Z","shell.execute_reply.started":"2022-10-07T16:25:55.88894Z","shell.execute_reply":"2022-10-07T16:25:55.906293Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n And with this, we are done with iteration 1! That was not as bad as it looked lol.
\n Now let's quickly go over through all steps of iteration 2.\n
","metadata":{}},{"cell_type":"code","source":"# iteration 2 (m = 2)\n\n# step 2A\ndf['r2'] = df['SalePrice'] - df['F1']\n\n# step 2B\nbaselearner2 = DecisionTreeRegressor(\n criterion='squared_error',\n splitter='best',\n min_samples_split=4\n)\nbaselearner2.fit(\n X=df[['NewConstruction', 'Pool', 'LotArea']],\n y=df['r2']\n)\ndf['gamma2'] = baselearner2.predict(X=df[['NewConstruction', 'Pool', 'LotArea']])\n\n# step2C\ndf['F2'] = df['F1'] + nu * df['gamma2']\n\ndf","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:36:57.701605Z","iopub.execute_input":"2022-10-07T16:36:57.702161Z","iopub.status.idle":"2022-10-07T16:36:57.736048Z","shell.execute_reply.started":"2022-10-07T16:36:57.702117Z","shell.execute_reply":"2022-10-07T16:36:57.734738Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n And with that, we are done with our iterations. Great!\n
","metadata":{}},{"cell_type":"markdown","source":"\n## Step 3","metadata":{}},{"cell_type":"markdown","source":"\n Step 3: $ F_M(x) $ will be our final predictor.\n
","metadata":{}},{"cell_type":"markdown","source":"\n Since we set $ M = 2 $ Base Estimators, our final predictor will be $ F_2(x) $, our F2 column in the dataset.\n
","metadata":{}},{"cell_type":"code","source":"df['F2']","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:42:30.721586Z","iopub.execute_input":"2022-10-07T16:42:30.723088Z","iopub.status.idle":"2022-10-07T16:42:30.734185Z","shell.execute_reply.started":"2022-10-07T16:42:30.723033Z","shell.execute_reply":"2022-10-07T16:42:30.733038Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n And with this, we are done going through our simple example!
\n Please keep in mind that in a real world setting, our number of Base Learners would be in the hundreds, thousands, or more, and our training dataset would not be comprised of only 5 rows.\n
","metadata":{}},{"cell_type":"markdown","source":"\n# π₯ XGBoost Overview","metadata":{}},{"cell_type":"markdown","source":"\n XGBoost is an implementation of Gradient Boosting with several extra features implemented to improve model performance and execution speed.\n
","metadata":{}},{"cell_type":"markdown","source":"\n Model Features:\n
\n - Standard Gradient Boosting algorithm.
\n - Stochastic Gradient Boosting with sub-sampling at the row, column and column per split levels.
\n - Regularized Gradient Boosting with both L1 and L2 regularization.
\n
\n
","metadata":{}},{"cell_type":"markdown","source":"\n System Features:\n
\n - Parallelization of tree construction using all of your CPU (or GPU) cores during training.
\n - Distributed Computing for training very large models using a cluster of machines.
\n - Out-of-Core Computing for very large datasets that donβt fit into memory.
\n - Cache Optimization of data structures and algorithm to make best use of hardware.
\n
\n
","metadata":{}},{"cell_type":"markdown","source":"\n Algorithm Features:\n
\n - Sparse Aware implementation with automatic handling of missing data values.
\n - Block Structure to support the parallelization of tree construction.
\n - Continued Training so that you can further boost an already fitted model on new data.
\n
\n
","metadata":{}},{"cell_type":"markdown","source":"\n# π» Code Implementation","metadata":{}},{"cell_type":"markdown","source":"\n xgboost's XGBRegressor will be doing all of the heavy lifting for us. Now that we have a better understanding of how it works, we can better appreciate open source development!\n
","metadata":{}},{"cell_type":"code","source":"# defining our tree object with the hyper parameters discussed above\nmodel = XGBRegressor(\n objective='reg:squarederror',\n n_estimators=200,\n learning_rate=0.1,\n)\n\n# XGBoost version of Gradient Boosting Algorithm that builds the tree based on our data\nmodel.fit(\n X=df[['NewConstruction', 'Pool', 'LotArea']],\n y=df['SalePrice']\n)","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:57:57.770259Z","iopub.execute_input":"2022-10-07T16:57:57.770739Z","iopub.status.idle":"2022-10-07T16:57:58.214824Z","shell.execute_reply.started":"2022-10-07T16:57:57.770702Z","shell.execute_reply":"2022-10-07T16:57:58.213677Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n So now, let's predict the SalePrice of a house with the following characteristics.
\n NewConstruction = 0, Pool = 0, LotArea = 10000. \n
","metadata":{}},{"cell_type":"code","source":"inputs = pd.DataFrame([{\"NewConstruction\": 0, \"Pool\": 0, \"LotArea\": 10000}])\nprediction = model.predict(inputs)\nprint(f\"XGBoost's predicted Sale Price: {prediction}\")","metadata":{"execution":{"iopub.status.busy":"2022-10-07T16:58:01.458167Z","iopub.execute_input":"2022-10-07T16:58:01.458748Z","iopub.status.idle":"2022-10-07T16:58:01.472269Z","shell.execute_reply.started":"2022-10-07T16:58:01.458689Z","shell.execute_reply":"2022-10-07T16:58:01.471191Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n# π Conclusion","metadata":{}},{"cell_type":"markdown","source":"\n With this, our Gradient Boosting and XGBoost notebook has come to an end.
\n If there are any mistakes, please point them out to me in the commets, this is a living and breathing document.
\n For a more detailed look at how to build an individual Decision Tree with the CART Algorithm, I recommend these 2 notebooks:
\n
\n For a related Ensemble Learning method:
π³ Random Forest\n
","metadata":{}},{"cell_type":"markdown","source":"","metadata":{}},{"cell_type":"markdown","source":"","metadata":{}},{"cell_type":"markdown","source":"\n Stay tuned for more notebooks in the ML Grandmaster series.
\n I hope you find this series of notebooks useful/fun!\n
","metadata":{}},{"cell_type":"markdown","source":"## ML Grandmaster Notebooks\n
\n","metadata":{}},{"cell_type":"markdown","source":"\n π Machine Learning Grandmaster\n
","metadata":{}}]}