Description
Excessive Efficiency Time Collection
Excessive Efficiency Time Collection
Turn out to be the time-series area skilled on your group
—————
Turn out to be the Time Collection Professional
on your group
The Excessive-Efficiency Time Collection Forecasting Course is a tremendous course designed to show Enterprise Analysts and Information Scientists find out how to scale back forecast error utilizing state-of-the-art forecasting methods which have received competitions. You’ll endure a full transformation studying essentially the most in-demand abilities that organizations want proper now. Time to speed up your profession.
—————
Crafted For Enterprise Analysts & Information Scientists:
That want to scale back forecasting error and scale outcomes on your group.
That is probably my most difficult course ever. You’ll be taught the time sequence abilities which have taken me 10-years of examine, follow, and experimentation.
My speak on Excessive-Efficiency Time Collection Forecasting
This course provides you the instruments it’s essential meet in the present day’s forecasting calls for.
A full 12 months was spent on constructing two of the software program packages you’ll be taught, modeltime and timetk.
Plus, I’m instructing you GluonTS, a state-of-the-art deep studying framework for time sequence written in python.
This course will problem you. It’ll change you. It did me.
– Matt Dancho, Course Teacher & Founding father of Enterprise Science
—————
Endure a Full Transformation
By studying forecasting methods that get outcomes
With Excessive-Efficiency Forecasting, you’ll endure an entire transformation by studying essentially the most in-demand abilities for creating high-accuracy forecasts.
By means of this course, you’ll be taught and apply:
✧ Machine Studying & Deep Studying
✧ Characteristic Engineering
✧ Visualization & Information Wrangling
✧ Transformations
✧ Hyper Parameter Tuning
✧ Forecasting at Scale (Time Collection Teams)
—————
The way it works:
Your path to changing into an Professional Forecaster is simplified into 3 streamlined steps.
1️⃣ Time Collection Characteristic Engineering
2️⃣ Machine Studying for Time Collection
3️⃣ Deep Studying for Time Collection
—————
►Half 1◄
Time Collection Characteristic Engineering:
First, we construct your time sequence characteristic engineering abilities. You be taught:
– Visualization: Figuring out options visually utilizing the simplest plotting methods
– Information Wrangling: Aggregating, padding, cleansing, and lengthening time sequence information
– Transformations: Rolling, Lagging, Differencing, Creating Fourier Collection, and extra
– Characteristic Engineering: Over 3-hours of content material on introductory and superior characteristic engineering
►Half 2◄
Machine Studying for Time Collection:
Subsequent, we construct your time sequence machine studying abilities. You be taught:
– 17 Algorithms: 8 hours of content material on 17 TOP Algorithms. Divided into 5 teams:
– ARIMA
– Prophet
– Exponential Smoothing – ETS, TBATS, Seasonal Decomposition
– Machine Studying – Elastic Internet, MARS, SVM, KNN, Random Forest, XGBOOST, Cubist, NNET & NNETAR
– Boosted Algorithms – Prophet Enhance & ARIMA Enhance
– Hyper Parameter Tuning: Methods to scale back overfitting & enhance mannequin efficiency
– Time Collection Teams: Scale your evaluation from one time sequence to a whole bunch
– Parallel Processing: Wanted to hurry up hyper parameter tuning and forecasting at scale
– Ensembling: Combining many algorithms right into a single tremendous learner
►Half 3◄
Deep Studying for Time Collection:
Subsequent, we construct your time sequence deep studying abilities. You be taught:
- GluonTS: A state-of-the-art forecasting bundle that’s constructed on high of mxnet (made by Amazon)
- Algorithms: Be taught DeepAR, DeepVAR, NBEATS, and extra!
Challenges & Cheat Sheets
Subsequent, we construct your time sequence machine studying abilities. You be taught:
– Cheat Sheets: Developed to make your forecasting workflow reproducible on any downside
– Challenges: Designed to check your talents & solidify your data
—————
Abstract of what you get
A methodical coaching plan that goes from idea to manufacturing ($10,000 worth)
- Half 1 – Characteristic Engineering with Timetk
- Half 2 – Machine Studying with Modeltime
- Half 3 – Deep Studying with GluonTS
- Challenges & Cheat Sheets
—————
Course Curriculum
Welcome to Excessive Efficiency Time Collection!
- Excessive-Efficiency Time Collection – Turn out to be the Time Collection Professional for Your Group (2:34)
- Personal Slack Channel – The best way to Be a part of
- Video Subtitles (Captions)
- What’s a Excessive-Efficiency Forecasting System?
- [IMPORTANT] System Necessities – R + Python Necessities & Widespread Points
Conditions
- Prerequisite – Information Science for Enterprise Half 1
Getting Assist
- Getting Assist (IMPORTANT!!!)
Module 0 – Introduction to Excessive-Efficiency Forecasting
- Excessive-Efficiency Forecasting – What You’re Studying, Why You’re Studying It (0:43)
0.1 Forecast Competitors Evaluation
- The Forecasting Competitors Evaluation & Course Development (3:34)
- 2014 Kaggle Walmart Recruiting Problem (5:11)
- 2018 M4 Competitors (3:37)
- 2018 Kaggle Wikipedia Web site Visitors Forecasting Competitors (4:30)
- 2020 M5 Competitors (5:59)
- 5 Key Takeaways from the Forecast Competitors Evaluation (5:41)
0.2 Course Tasks – Google Analytics, E-mail Subscribers, & Gross sales Forecasting
- The Enterprise Case – Growing a Greatest-in-Class Forecasting System (3:03)
0.3 What Instruments are in Your Toolbox?
- Timetk: Time Collection Information Preparation, Visualization, & Preprocessing (5:54)
- Modeltime: Time Collection Machine Studying (5:25)
- GluonTS: Time Collection Deep Studying (2:01)
- [Cheat Sheet] Forecasting Workflow
Module 01 – Time Collection Bounce
- Time Collection Bounce (0:54)
1.1 Time Collection Venture Setup
- Venture Setup (2:28)
- Course Information (File Obtain) (1:02)
- R Package deal Set up – Half 1 (File Obtain) (5:26)
- R Package deal Set up – Half 2 (5:14)
- Bounce Setup (File Obtain) (0:44)
1.2 Enterprise Understanding & Dataset Terminology
- Set up Relationships, Half 1 – Google Analytics Abstract Dataset (4:11)
- Set up Relationships, Half 2 – Google Analytics High 20 Pages (5:23)
- Construct Relationships – Mailchimp & Studying Lab Occasions (4:49)
- Generate Course Income – Transaction Income & Product Occasions (3:03)
- Code Checkpoint (File Obtain) (0:54)
1.3 TS Bounce: Dive into Forecasting E-mail Subscribers!
- Learn This! – Time Collection Bounce Intent
- Time Collection Bounce – Setup (File Obtain) (3:20)
- Libraries & Information (3:13)
1.3.1 Exploratory Information Evaluation for Time Collection
- EDA for Time Collection (1:08)
- Summarize By Time (5:46)
- Time Collection Abstract Diagnostics (4:47)
- Pad by Time (4:08)
- Visualize the Time Collection (3:12)
1.3.2 Analysis & Prepare/Check Home windows
- Analysis Window – Filter By Time (4:43)
- Time Collection Prepare/Check Break up (4:53)
1.3.3 Forecasting with Prophet
- Coaching a Prophet Mannequin with Modeltime (4:21)
- Modeltime Forecasting Workflow – Spherical 1 (7:43)
1.3.4 Forecasting with Characteristic Engineering
- Visualizing Seasonality (4:34)
- Characteristic Engineering – Half 1 (5:45)
- Characteristic Engineering – Half 2 (5:51)
- Machine Studying with Workflows (3:35)
- Modeltime Forecasting Workflow – Spherical 2 (5:59)
1.3.5 Recap & Code Checkpoint – Module 01 – TS Bounce
- Right here’s the place you’re going. (3:11)
- Code Checkpoint (File Obtain)
[Part 1] Time Collection with Timetk
- Welcome to Half 1 – Time Collection with Timetk! (2:17)
Module 02 – Time Collection Visualization
- Setup (File Obtain) & Overview – Visualization (2:11)
- Information Preparation – Half 1 (4:29)
- Information Preparation – Half 2 (3:23)
2.1 Time Collection Plots [MUST KNOW FUNCTION]
- [MUST KNOW] Plotting Time Collection (5:31)
- Plotting with Transformations (4:37)
- Adjusting the Smoother (6:11)
- Smoother for Teams (1:54)
- Interactive & Static Plots (2:00)
2.2 Autocorrelation Plots
- ACF & PACF Ideas – Autocorrelation & Partial Autocorrelation
- ACF & PACF Plotting (7:49)
- Lag Adjustment (1:24)
- CCF Plotting – Cross Correlations (7:58)
2.3 Seasonality Plots
- Seasonality Field Plot (5:52)
- Seasonality Violin Plot (0:53)
2.4 Anomaly Plots
- Anomaly Plot Fundamentals (4:50)
- Getting the Anomaly Information (2:00)
- Working with Grouped Information (1:43)
2.5 STL Decomposition & Regression Plots
- STL Decomposition Plot (4:44)
- STL Decomposition – Grouped Time Collection (2:11)
2.6 Regression Plots [SECRET WEAPON FOR FEATURE ENGINEERING]
- [SECRET WEAPON] Time Collection Regression Plot (7:08)
- Time Collection Regression Plot – Grouped Time Collection (4:05)
2.7 Code Checkpoint – Module 02 – Visualization
- Code Checkpoint (File Obtain)
Module 03 – Time Collection Information Wrangling
- Setup (File Obtain) & Overview – Information Wrangling (2:34)
3.1 Summarise By Time [MUST KNOW]
- Single & Grouped Time Collection Summarizations (4:37)
- Utilizing Throughout (to Summarize Broad-Format Tibbles by Time) (5:11)
- Weekly/Month-to-month/Quarterly/Yearly Aggregations (3:33)
- Ground, Ceiling, Spherical (5:15)
3.2 Pad by Time
- Filling in Gaps (2:54)
- From Low-Frequency to Excessive-Frequency (3:36)
3.3 Filter By Time
- Zooming & Slicing (5:14)
- Offsetting by Time (2:01)
3.4 Mutate By Time
- Extrapolate the Imply, Median, Max, Min By Time (7:57)
3.5 Becoming a member of By Time
- Combining Subscribers & Internet Visitors (3:48)
- Inspecting the Be a part of (3:00)
- Formatting the Be a part of for Characteristic Relationships (5:49)
- Be a part of Cross Correlations (3:22)
3.6 Time Collection Index Operations
- Making a Time Collection (4:39)
- Making a Vacation Sequence (3:14)
- Time Offsets (3:01)
- Making a Future Time Collection (3:12)
3.7 Forecasting with Future Frames
- The Future Body (2:47)
- [FORECAST SPOTLIGHT] Forecasting with the Future Body 📈 (6:53)
3.8 Code Checkpoint – Module 03 – Information Wrangling
- Code Checkpoint (File Obtain)
Module 04 – Transformations for Time Collection
- Setup (File Obtain) & Overview – Transformations (2:15)
- Libraries & Information (2:12)
4.1 Variance Discount Transformations – Log & Field Cox [MUST KNOW]
- Why is Variance Discount Vital? (4:43)
- Log – Log (and Log1P) Transformation (4:17)
- Log – Assessing the Advantage of Log1P Transformation (2:51)
- Log – Teams & Inversion (3:43)
- Field Cox – What’s the Field Cox Transformation? (2:34)
- Field Cox – Assessing the Profit (4:04)
- Field Cox – Inversion (2:05)
- Field Cox – Managing Grouped Transformations & Inversion (8:36)
4.2 Rolling & Smoothing Transformations
- Introduction to Rolling & Smoothing (1:49)
- Rolling Home windows – What’s a Transferring Common? (File Obtain) (3:53)
- Rolling Home windows – Transferring Common & Median Utilized (8:53)
- Loess Smoother (7:02)
- Rolling Correlation – Slidify, Half 1 (4:16)
- Rolling Correlation – Slidify, Half 2 (7:40)
- [BUSINESS SPOTLIGHT] The Drawback with Forecasting utilizing a Transferring Common (6:43)
4.3 Vary Discount Transformations
- Introduction to Normalization & Standardization (0:59)
- What’s Normalization? [Min = 0, Max = 1] (4:50)
- What’s Standardization? [Mean = 0, Standard Deviation = 1] (2:31)
4.4 Imputation & Outlier Cleansing
- Introduction to Imputation & Outlier Cleansing (0:44)
- Imputation – Time Collection NA Restore (6:40)
- Anomalies – Time Collection Outlier Cleansing (7:22)
- Anomalies – When to Take away Outliers (5:21)
4.5 Lags & Differencing Transformations [MUST KNOW]
- Introduction to Lags & Differencing (1:08)
- Lags – What’s a Lag? (1:49)
- Lags – Lag Detection with ACF/PACF (3:54)
- Lags – Regression with Lags (5:06)
- Differencing – Progress vs Change (4:00)
- Differencing – Acceleration (6:22)
- Differencing – Evaluating A number of Time Collection (4:44)
- Differencing – Inversion (0:57)
4.6 Fourier Collection [MUST KNOW]
- Introduction to the Fourier Collection (7:23)
- Fourier Regression (4:24)
4.7 Constrained Interval Forecasting [FORECAST SPOTLIGHT] 📈
- What’s the Log Interval Transformation? (5:47)
- Visualizing the Transformation (4:12)
- Transformations & Preprocessing (5:09)
- Modeling (6:29)
- Making ready Future Information (3:36)
- Making Predictions (1:05)
- Combining the Forecast Information (4:08)
- Estimating Confidence Intervals (8:24)
- Visualizing Confidence Intervals (2:10)
- Inverting the Log Interval Transformation (4:08)
4.8 Code Checkpoint – Module 04 – Transformations
- Code Checkpoint (File Obtain)
Problem #1 – Exploring Transactions & Internet Web page Visitors
- Problem #1 Dialogue (File Obtain) (4:21)
- Answer – Half 1 (File Obtain) (7:18)
- Answer – Half 2: Begins at “Establish Relationships” (7:51)
Module 05 – Introduction to Characteristic Engineering (for Time Collection)
- Setup (File Obtain) & Overview – Intro to Characteristic Engineering (2:30)
- Information Prep, Half 1 – Log Standardize (5:27)
- Information Prep, Half 2 – Getting Able to Clear (5:01)
- Information Prep, Half 3 – Focused Cleansing with Between Time (4:18)
5.1 Time-Primarily based Options (Development & Seasonal/Calendar) [MUST KNOW]
- The Time Collection Signature (7:55)
- Characteristic Removing (3:28)
- Linear Development (2:10)
- Non-Linear Development – Foundation Splines (4:41)
- Non-Linear Development – Pure Splines (Stiffer than Foundation Splines) (4:29)
- Seasonal Options – Weekday & Month (3:21)
- Seasonal Options – Combining with Development (5:23)
5.2 Interactions
- Interplay Options – Spikes Each Different Wednesday (7:35)
5.3 Fourier Options
- Deciding on & Including Fourier Frequency Options (4:21)
- Modeling & Visualizing the Fourier Results (2:07)
5.4 Autocorrelated Lag Options
- Deciding on & Including Lag Options (6:59)
- Modeling & Visualizing the Lag Results (5:20)
5.5 Particular Occasion Options
- Making ready Occasion Information for Evaluation (6:34)
- Visualizing Occasions (2:57)
- Modeling & Visualizing Occasion Results (2:08)
- Fixing the Spline (2:07)
5.6 Exterior Regressors (Xregs)
- Reworking Xregs (5:05)
- Becoming a member of Xregs (1:49)
- Inspecting Cross Correlations (1:53)
- Modeling with Xregs (3:28)
- Visualizing PageViews vs Optins & Modeling Lags (6:58)
5.7 Beneficial Mannequin Options
- Amassing the Beneficial Mannequin (3:44)
- Saving the Mannequin Artifact (2:28)
5.8 Code Checkpoint – Module 05 – Introduction to Characteristic Engineering
- Code Checkpoint (File Obtain)
Module 06 – Superior Characteristic Engineering Workflow
- Forecasting Workflow [CHEAT SHEET] (3:40)
- Setup (File Obtain) & Overview – Superior Characteristic Engineering (1:43)
- Information Preparation (4:42)
6.1 Creating the “Full” Dataset – Extending & Including Lagged Options & Occasions [IMPORTANT]
- The “Full” Dataset (2:50)
- Extending – Future Body (3:21)
- Including Lag Options (4:02)
- Add Lagged Rolling Options (5:03)
- Add Occasions (Exterior Regressors) (2:57)
- Format Column Names (3:09)
6.2 Separate into Modeling Information & Forecast Information
- Information Ready / Future Information Break up (2:48)
6.3 Separate into Coaching Information & Testing Information
- Prepare / Check Break up (3:55)
6.4 Recipes – Characteristic Engineering Pipeline Steps
- Recipes Intro (2:41)
- Step – Time Collection Signature Options (5:48)
- Step – Characteristic Removing (3:10)
- Step – Standardization (2:11)
- Step – One-Sizzling Encoding (1:55)
- Step – Interplay Options (2:28)
- Step – Fourier Collection Options (2:03)
6.5 Constructing the Spline Mannequin
- Mannequin Spec: LM Mannequin (1:02)
- Recipe Spec: Spline Options (5:59)
- Workflow: Spline Recipe + LM Mannequin (2:49)
6.6 Introduction to Modeltime Workflow
- Modeltime Desk & Calibration (2:08)
- Forecasting the Check Information (2:40)
- Measuring the Check Accuracy (1:19)
- Evaluating the Coaching & Testing Accuracy (1:32)
6.7 Constructing the Lag Mannequin
- Recipe Spec: Lag Options (3:00)
- Workflow: Lag Recipe+ LM Mannequin (2:40)
- Modeltime: Evaluating Spline & Lag Fashions (4:23)
6.8 Forecasting the Future
- Refitting the Fashions (4:37)
- Transformation Inversion (5:23)
- Visualizing the Forecast within the Unique Scale (1:59)
6.9 Saving the Artifacts
- Creating an Artifact Record, Half 1 (4:34)
- Creating an Artifact Record, Half 2 (3:11)
- Organizing the Artifacts Record (1:57)
- Saving the Artifacts (1:28)
6.10 Code Checkpoint – Module 06 – Superior Characteristic Engineering
- Code Checkpoint (File Obtain)
Problem #2 – Characteristic Engineering & Modeltime Workflow [YOU’VE GOT THIS!]
- Problem Dialogue, Half 1 (File Obtain) – Characteristic Preparation (5:11)
- Problem Dialogue, Half 2 – Characteristic Engineering & Modeling (4:56)
Problem #2 – Answer
- Answer, Half 1 (File Obtain) – Accumulate & Put together Information (3:49)
- Answer, Half 2 – Visualizations (3:19)
- Answer, Half 3A – Create Full Dataset (5:46)
- Answer, Half 3B – Visualize the Full Dataset (3:47)
- Answer, Half 4 – Mannequin/Forecast Information Break up (1:05)
- Answer, Half 5 – Prepare/Check Information Break up (0:56)
- Answer, Half 6 – Characteristic Engineering (4:18)
- Answer, Half 7 – Modeling: Spline Mannequin (6:08)
- Answer, Half 8 – Modeling: Lag Mannequin (2:25)
- Answer, Half 9 – Modeltime (4:03)
- Answer, Half 10 – Forecast (6:49)
Problem #2 Bonus – Regularization
- Regularization, Half 1 (File Obtain) – Mannequin: GLMnet (4:01)
- Regularization, Half 2 – Enhancing the Lag Mannequin with GLMNet (5:28)
- Regularization, Half 3 – Forecasting the Future Information with GLMNet + Lag Recipe (3:02)
Half 1 Full – You rock!
- WOOO HOOO – You crushed it!
[Part 2] Machine Studying for Time Collection with Modeltime
- Choosing Up From Half 1 (Venture Obtain)
Module 07 – Modeltime Workflow [DEEP DIVE]
- Setup – Modeltime Workflow [In-Depth] (1:25)
- Overview – Modeltime Workflow [In-Depth] (1:16)
- Libraries & Artifacts Preparation (2:33)
7.1 Making Fashions – Object Sorts & Necessities
- Mannequin Necessities for Modeltime (1:34)
- Parsnip Object Fashions – Univariate (3:37)
- Workflow Objects – Multivariate, Date-Primarily based Options (7:14)
- Workflow Object – Multivariate, Exterior Options (4:53)
7.2 Modeltime Desk
- Modeltime Desk – Key Necessities (4:27)
7.3 Calibration Desk
- Calibration Desk – How It Works (3:29)
7.4 Measuring Mannequin Accuracy [IMPORTANT!!!]
- Major Accuracy Metrics & Makes use of [SUPER IMPORTANT] (7:40)
- Customized Metric Units utilizing Yardstick (3:54)
- Customizing the Accuracy Desk Output (3:28)
7.5 Forecasting the Check Information
- Modeltime Forecast – How It Works (6:22)
- Customizing the Forecast Visualization (5:00)
7.6 Mannequin Refitting & Forecasting
- Refitting – How It Works (3:02)
- Making the Forecast (5:20)
7.7 Code Checkpoint – Module 07A – Modeltime Workflow [In-Depth]
- Code Checkpoint (File Obtain)
7.8 New Options of Modeltime 0.1.0 – Module 07B
- Setup (File Obtain) – Modeltime New Options (1:53)
- Expedited Forecasting – Modeltime Desk (5:20)
- Expedited Forecasting – Skip Straight to Forecasting (2:20)
- Visualizing a Fitted Mannequin (2:57)
- Calibration – In-Pattern vs Out-of-Pattern Accuracy (5:25)
- Residual Diagnostics – Getting Residuals (2:16)
- Residuals – Time Plot (2:39)
- Residuals – Plot Customization (2:29)
- Residuals – ACF Plot (4:06)
- Residuals – Seasonality Plot (3:50)
7.9 Code Checkpoint – Module 07B – Modeltime New Options!
- Code Checkpoint (File Obtain)
Module 08 – ARIMA
- Setup (File Obtain) (0:40)
- ARIMA Coaching Overview (1:29)
- Libraries & Artifacts Setup (1:49)
8.1 ARIMA Ideas
- Auto-Regressive Features: ar() & arima() (5:15)
- Auto-Regressive (AR) Modeling with Linear Regression (3:11)
- Single-Step Forecast for AR Fashions (4:43)
- Multi-Step Recursive Forecasting for AR Fashions (4:44)
- Integration (Differencing) (5:42)
- Transferring Common (MA) Course of (Error Modeling) (7:36)
- Seasonal ARIMA (SARIMA) (4:29)
- Including XREGS (SARIMAX) (4:44)
8.2 ARIMA in Modeltime
- Setting Up Fundamental ARIMA in Modeltime (4:45)
- Making an attempt Totally different ARIMA Parameters (5:11)
- About AIC (Akaike Data Criterion) (3:42)
8.3 Modeltime Auto ARIMA
- Implementing Auto ARIMA in Modeltime (1:49)
- How Auto ARIMA Works – Lazy Grid Search (1:27)
- Evaluating ARIMA & Auto ARIMA (3:15)
- Including Fourier Options to Choose Up Greater than 1 Seasonality (3:49)
- Including Occasion Options to Enhance R-Squared (Variance Defined) (1:33)
- Refitting & Reviewing the Forecast (2:57)
- Including Month Options to Account for February Enhance – BEST MAE 0.564 (3:35)
8.4 Recap – ARIMA
- ARIMA Strengths & Weaknesses (and Methods that Labored) (3:56)
- Saving Artifacts – Greatest ARIMA Mannequin (3:28)
8.5 Code Checkpoint – Module 08 – ARIMA
- Code Checkpoint (File Obtain)
Module 09 – Prophet
- Setup (File Obtain) (0:27)
- Prophet Coaching Overview (0:51)
- Libraries & Artifacts (2:02)
9.1 Prophet with Modeltime
- Prophet Regression: prophet_reg() (3:23)
- Modeltime Workflow (2:02)
- Adjusting the Key Prophet Parameters (5:13)
9.2 Prophet Ideas
- Extracting the Prophet Mannequin from Modeltime (3:11)
- Visualizing the Impact of Key Parameters on the Prophet Mannequin (5:48)
- Understanding Prophet Parts & Additive Mannequin (2:37)
9.3 Again to Modeling with Prophet – XREGS!
- Becoming Prophet w/ Occasions (2:19)
- Evaluating No Occasions vs Occasions – BEST MAE 0.488 (w/ Occasions) 🚀 (3:05)
- Making the Forecast (2:10)
9.4 Recap – Prophet
- Logging (Saving) Your Progress (2:40)
- Recap – Prophet Strengths & Weaknesses (3:02)
9.5 Checkpoint – Module 09 – Prophet
- Code Checkpoint (File Obtain)
Module 10 – Exponential Smoothing, TBATS, & Seasonal Decomposition
- Setup (File Obtain) (0:18)
- Overview – Exponential Smoothing (0:35)
- Libraries & Artifacts (1:37)
10.1 Exponential Smoothing
- The Exponential Weighting Operate (4:50)
- Making use of the Exponential Weighting Operate to Make a Forecast (2:41)
- ETS Mannequin: exp_smoothing() (3:52)
- Visualizing the ETS Mannequin (4:48)
10.2 TBATS
- TBATS Mannequin: seasonal_reg() (3:36)
- Visualizing the TBATS Mannequin (2:48)
10.3 Seasonal Decomposition Fashions
- Seasonal Decomposition & A number of Seasonality Time Collection (MSTS) Objects (2:28)
- STLM ETS Mannequin (2:33)
- STL Plot & Relationship to STLM ETS Mannequin (2:49)
- STLM ARIMA Mannequin (1:55)
- STLM ARIMA – Including XREGS (1:08)
10.4 Analysis
- Making ready the Check Forecast Visualization (3:30)
- Evaluating A number of Fashions – ETS, TBATS, STLM ARIMA & ETS – BEST MAE 0.523 (TBATS) (3:45)
- Refitting – Inspecting the Future Forecasts (3:34)
10.5 Recap – ETS, TBATS, Seasonal Decomp
- Saving Artifacts (2:22)
- Strengths & Weaknesses – ETS, TBATS, Seasonal Decomp (2:05)
10.6 Code Checkpoint – Module 10 – ETS, TBATS, & Seasonal Decomposition
- Code Checkpoint (File Obtain)
Problem #3 – ARIMA + Prophet + ETS + TBATS
- Problem #3 Dialogue, Half 1 (File Obtain) – by means of ARIMA (5:32)
- Problem #3 Dialogue, Half 2 – Prophet to Finish of Problem (2:33)
Problem #3 – Answer
- Answer, Half 1 – Prepare/Check Setup (Answer File Obtain) (1:55)
- Answer, Half 2 – ARIMA (Mannequin 1): Fundamental Auto ARIMA (3:03)
- Answer, Half 3 – ARIMA (Mannequin 2): Auto ARIMA + Including Product Occasions (2:14)
- Answer, Half 4 – ARIMA (Mannequin 3): Auto ARIMA + Occasions + Seasonality (2:08)
- Answer, Half 5 – ARIMA (Mannequin 4): Forcing Seasonality with Guide ARIMA (1:17)
- Answer, Half 6 – ARIMA (Mannequin 5): Auto ARIMA + Occasions + Fourier Collection (0:57)
- Answer, Half 7 – ARIMA – Modeltime Workflow (2:26)
- Answer, Half 8 – ARIMA – Forecast Evaluation (3:18)
- Answer, Half 9 – Prophet Fashions: Fundamental (6), Yearly Seasonality (7), Occasions (8), Occasions + Fourier (9) (2:52)
- Answer, Half 10 – Prophet – Modeltime Workflow (1:38)
- Answer, Half 11 – Prophet – Forecast Evaluation (3:13)
- Answer, Half 12 – Exponential Smoothing Fashions: ETS (10), TBATS (11) (3:24)
- Answer, Half 13 – Exponential Smoothing – Modeltime Workflow (1:45)
- Answer, Half 14 – Exponential Smoothing – Forecast Evaluation (1:30)
- Answer, Half 15 – Forecasting the Future Information – ARIMA, Prophet & ETS/TBATS (3:40)
- Answer, Half 16 – Ultimate Evaluation – ARIMA, Prophet, & ETS/TBATS (2:47)
Problem #3 BONUS – ARIMA & Prophet vs Linear Mannequin
- Bonus, Half 1 (File Obtain) – Including the LM from Problem #2 (4:43)
- Bonus, Half 2 – Why is the LM forecast excessive in March? (4:41)
11.0 Machine Studying Algorithms [IMPORTANT]
- Welcome to Machine Studying for Time Collection (File Obtain) (5:22)
11.1 Elastic Internet Algorithm (GLMNet) – Linear
- GLMNet – Mannequin Spec (3:43)
- GLMNet – Spline & Lag Workflows (2:40)
- GLMNet – Calibration, Accuracy, & Plot (4:06)
- GLMNet – Tweaking Parameters – BEST MAE 0.519 (Lag Mannequin) (2:33)
Plotting Utility – Let’s make a helper operate to hurry analysis up!
- calibrate_and_plot() (5:50)
- Visualizing the Impact of Parameter Changes (3:19)
11.2 A number of Adaptive Regression Splines (MARS) – Linear
- We come from MARS (3:30)
- MARS – A Easy Instance (6:55)
- MARS – Spline & Lag Fashions – BEST MAE 0.518 (Spline Mannequin) (4:28)
11.3 Assist Vector Machine (SVM) – Polynomial
- SVM Polynomial – Mannequin Specification (2:54)
- SVM Poly – Tweaking Parameters – BEST MAE 0.615 (Spline Mannequin) – BOOO (5:09)
11.4 Assist Vector Machine (SMV) – Radial Foundation Operate
- 16% Enchancment – SVM RBF vs SVM Poly (2:29)
- SVM RBF – Parameter Tweaking (3:11)
- SVM RBF – Lag Mannequin – BEST MAE 0.520 (Spline Mannequin) – Niiiice! (1:55)
11.5 [Important Concept] KNN & Tree-Primarily based Algorithms – The Drawback with Predicting Time Collection Traits
- Strengths/Weak spot – KNN & Tree-Primarily based Algorithms Can’t Predict Past the Min/Max (1:24)
- KNN vs GLMNET – Making Pattern Information with Development (2:08)
- KNN vs GLMNET – Making Easy Development Fashions (4:12)
- KNN vs GLMNET – Visualize the Development Predictions w/ Modeltime – Yikes, GLMNET simply schooled KNN (4:14)
11.5 Ok-Nearest Neighbors (KNN) – Similarity (Distance) Primarily based
- KNN – Spline Mannequin (3:30)
- KNN – Tweaking Key Parameters (5:52)
- KNN – Lag Mannequin – BEST MAE 0.558 (Spline Mannequin) (2:05)
You’re kicking butt… However, don’t neglect to take breaks
- [COFFEE BREAK] With Invoice Murray
11.6 Random Forest (Tree-Primarily based)
- RF – Spline Mannequin (4:27)
- RF – Lag Mannequin – 32% Higher vs Spline Mannequin (3:11)
- RF – Tweaking Parameters – BEST MAE 0.516 (Lag Mannequin) (4:02)
11.7 XGBoost (Gradient Boosting Machine) – Tree-Primarily based
- XGBoost – Spline & Lag Fashions (5:00)
- XGBoost – Tweaking Parameters – 0.484 MAE (Lag Mannequin) (6:35)
- XGBoost – Tweaking Parameters 2 – BEST MAE 0.484 (Lag Mannequin) 🚀 (3:32)
11.8 Cubist – Combo of Timber (Guidelines) + Linear Fashions at Nodes
- Cubist – Spline & Lag Fashions – 0.514 MAE out of the gate! (4:53)
- Cubist – Tweaking Parameters – OPTIMAL MAE / R-SQUARED (0.524 / 0.316) (5:48)
11.9 Neural Internet (NNET) – Like a Linear Regression however Higher
- NNET – Spline & Lag Fashions (4:57)
- NNET – Tweaking Parameters – BEST MAE 0.553 (Spline Mannequin) (5:39)
11.10 NNETAR – Combining AR Phrases with a NNET!
- What the heck is NNETAR? (NNET + ARIMA – IMA = NNETAR) (2:22)
- NNETAR – Mannequin, Recipe, & Workflow (4:11)
- NNETAR – Tweaking AR Parameters (2:24)
- NNETAR – Tweaking NNET Parameters – BEST MAE 0.512 (4:13)
11.11 Modeltime Experimentation Evaluation
- Organizing in a Modeltime Desk (4:22)
- Updating the Descriptions Programmatically (4:02)
- Mannequin Choice – Course of & Ideas (utilizing Accuracy Desk) (3:39)
- Mannequin Inspection – Course of & Ideas (utilizing Check Forecast Visualization) (3:03)
- Mannequin Inspection – Visualizing the Future Forecast (5:42)
11.12 Saving Your Work – Artifacts!
- Saving Fashions (2:34)
- Saving your calibrate_and_plot() operate (1:29)
11.13 Checkpoint – Module 11 – Machine Studying Algorithms
- Code Checkpoint (File Obtain)
12.0 Boosted Algorithms – Prophet Enhance & ARIMA Enhance
- Boosted Algorithms – A Highly effective Approach for Enhancing Efficiency (3:37)
12.1A Prophet Baseline Mannequin
- Baseline: Greatest Prophet Mannequin (2:38)
- [Pro Tip] The best way to Repair a Damaged Mannequin (2:50)
- Prophet Baseline – Greatest Mannequin MAE 0.488 (0:54)
12.1B Prophet Enhance
- Recipe for Prophet Enhance (3:33)
- Mannequin Technique – Utilizing XGBOOST for Seasonality/XREG Modeling (4:39)
- Workflow – No Parameter Tweaking (3:41)
- [KEY CONCEPT] Prophet Enhance – Modeling Development with Prophet, Residuals with XGBoost (3:00)
- Prophet Enhance – Tweaking Parameters – BEST MAE 0.457 🚀 (6:33)
12.2 ARIMA Enhance
- Modeling Technique – ARIMA for pattern, XGBOOST for XREGS (3:50)
- ARIMA Enhance – Mannequin Specification (5:57)
- ARIMA Enhance – Tweaking Parameters – BEST MAE 0.523 (4:34)
12.3 Boosted Fashions – Modeltime Workflow
- Modeltime – Accuracy Analysis & Figuring out Damaged Fashions (2:43)
- Modeltime – Forecast Check Information (2:10)
- Modeltime – Refitting & Forecasting Future (3:08)
- Save Your Work (1:26)
12.4 Code Checkpoint – Boosted Algorithms
- Code Checkpoint (File Obtain)
13.0 Hyper Parameter Tuning & Cross Validation – For Time Collection
- Hyperparameter Tuning for Time Collection (File Downloads) (3:56)
- [CHEAT SHEET] Hyperparameter Tuning Workflow (4:47)
- Getting ed – Setup & Workflow (3:09)
13.1 Reviewing 28 Fashions (It’s Simple with Modeltime)
- Combining Our Artifacts – 28 Fashions!(3:06)
- Accuracy Evaluation & Hyperparameter Tuning Candidate Choice (This Used to Take Me Weeks To Do) (4:36)
13.2 [SEQUENTIAL MODELS] NNETAR – Hyperparameter Tuning Course of
- What are Sequential Fashions? (& Why do we have to tune them otherwise?) (2:55)
- Extracting the Workflow from a Modeltime Desk: pluck_modeltime_model() (1:40)
- Time Collection Cross Validation (TSCV) Specification, Half 1: time_series_cv() (4:34)
- Time Collection Cross Validation (TSCV), Half 2: plot_time_series_cv_plan() (4:14)
- Establish Tuning Parameters – Recipe Spec (3:07)
- Establish Tuning Parameters – Mannequin Spec (5:14)
- Make a Grid for Parameters – Grid Spec (5:55)
13.2.1 – NNETAR Tuning, Spherical 1 – Default Params
- Grid Latin Hypercube Specification: grid_latin_hypercube() (3:19)
- Tuning Workflow Preparation (3:30)
- Tune Grid & Present Outcomes (7:24)
- Visualize the Parameter Outcomes (3:24)
13.2.2 NNETAR Tuning, Spherical 2 – Discovering the Candy Spot!
- Replace Grid Parameter Ranges (8:13)
- Parallel Processing – Pace-Up Tuning (5:13)
- Pace Comparability (Parallel vs Collection) – 3.4X Pace Enhance (44 sec vs 151 sec)
- Evaluation Parameters vs Efficiency Metrics (1:09)
- NNETAR – Prepare the Ultimate Mannequin – Greatest RMSE 0.507(4:15)
13.3 [NON-SEQUENTIAL MODELS] Prophet Enhance – Hyperparameter Tuning Course of
- What are Non-Sequential Fashions? (2:44)
- Mannequin Extraction: pluck_modeltime_model() (1:04)
- Ok-Fold Cross Validation (Use with Non-Sequential Fashions ONLY) (4:23)
- Prophet Enhance – Recipe (1:10)
- Prophet Enhance – Mannequin Spec (Establish Parameters for Tuning) (3:57)
13.3.1 Prophet Enhance Tuning, Spherical 1 – Default Parameters
- Grid Specification – Grid Latin Hypercube w/ Default Parameters (4:52)
- Tuning the Grid (in Parallel) (6:18)
- Visualize Outcomes – Studying Fee Dominates (2:58)
13.3.2 Prophet Enhance Tuning, Spherical 2 – Controlling Studying Fee
- Grid Specification – Controlling Studying Fee (4:45)
- Hyperparameter Tuning – Spherical 2 – We will see parameter tendencies!(3:17)
13.3.3 Prophet Enhance Tuning, Spherical 3 – Honing In
- Grid Specification & Tuning – Honing the parameter ranges in (5:49)
- Greatest RMSE Mannequin (Central Tendency) – MAE 0.466, RMSE 0.630, RSQ 0.450 (6:13)
- Greatest R-Squared Mannequin (Variance Defined) – MAE 0.464, RMSE 0.643, RSQ 0.459 (2:42)
13.4 Saving Our Progress
- Recap & Saving the Fashions (6:53)
13.5 Code Checkpoint – Mannequin 13 – Hyperparameter Tuning
- Code Checkpoint (File Obtain)
14.0 Ensemble Time Collection Fashions (Stacking)
- Competitors Ensembling Evaluation (5:57)
- What’s an Ensemble Mannequin? (7:21)
- Modeltime Ensemble: Documentation (2:01)
- Forecasting Cheat Sheet Improve [Download Here] (1:00)
14.1 Mannequin Efficiency Evaluation
- Code Setup [File Download] (6:49)
- Reviewing Fashions – Combining Tables & Organizing Outcomes (4:24)
- Reviewing Fashions – Making Sub-Mannequin Picks (7:46)
14.2 Common Ensemble
- Imply Ensemble – RMSE 0.640 vs 0.630 (Greatest Submodel) (5:00)
- Median Ensemble – RMSE 0.648 vs 0.630 (Greatest Submodel) (2:23)
14.3 Weighted Common Ensembles
- Introduction to Weighted Ensembles (1:02)
- Loading Choice (4:29)
- Accuracy Evaluation – RMSE 0.628 vs RMSE 0.630 (Baseline) (2:37)
14.4.A Stacked Ensembles – Stacking Course of
- Introduction to Meta-Learner Ensembling with Modeltime Ensemble (3:57)
- Resampling: Time Collection Cross Validation (TSCV) Technique (5:17)
- Making Sub-Mannequin CV Predictions – modeltime_fit_resamples() (4:27)
- Resampling & Sub-Mannequin Prediction: Ok-Fold Technique (6:28)
- Linear Regression Stack – TSCV – RMSE 1.00 (Ouch!) (7:16)
- Linear Regression Stack – Ok-Fold – RMSE 0.651 (A lot Higher, however We Can Do Higher)(3:25)
14.4.B Stacked Ensembles – Stacking with Tunable Algorithms
- GLMNET Stack – RMSE 0.641 (Heading in the right direction) (6:38)
- Modeltime Ensemble – In-Pattern Prediction Error – Bug Squashed (1:10)
- Random Forest Stack – RMSE 0.587!!! (7% enchancment) (4:33)
- Neural Internet Stack – RMSE 0.643 (4:05)
- XGBoost Stack – RMSE 0.585!!! (4:29)
- Cubist Stack – RMSE 0.649 (3:11)
- SVM Stack – RMSE 0.608!! (3:26)
14.5 Multi-Degree Stacking
- Degree 2 – Mannequin Analysis & Choice (4:27)
- Degree 3 – Weighted Ensemble Creation, Analysis, & Choice – RMSE 0.595 (Degree 2 RF is New Baseline RMSE 0.585) (3:34)
14.6 Modeltime Workflow for Ensembles
- Ensemble Calibration (4:45)
- Ensemble Refitting, Technique 1: Retraining Submodels Solely (5:43)
- Ensemble Refitting, Technique 2: Retraining each Sub-Fashions & Tremendous-Learners (5:33)
14.7 Saving Your Work
- Save the Multi-Degree Ensemble (1:27)
- Object Measurement: 50MB! Right here’s why. (3:15)
14.8 Code Checkpoint – Module 14 – Ensemble Strategies
- Code Checkpoint [File Download]
15.0 Forecasting at Scale – Time Collection Teams [Panel Data]
- Welcome to Module 15 – Forecasting at Scale utilizing Panel Information (Non-Recursive) Methods (2:30)
- Setup [File Download] (4:30)
15.1 Information Understanding & Preparation
- Information Understanding (4:33)
- Information Prep, Half 1: Padding by Group | Ungrouped Log Transformation (3:53)
- Information Prep, Half 2: Prolong by Group (2:44)
- Information Prep, Half 3: Fourier Options & Lag Options by Group (6:03)
- Information Prep, Half 4: Rolling Options by Group | Including a Row ID (4:59)
- Future & Ready Information – Preparation (7:34)
15.2 Time Splitting – Prepare/Check
- Time Collection Break up (Prepare/Check) (3:50)
15.3 Preprocessing & Recipes
- Cleansing Outliers by Group (5:18)
- Recipe, Half 1: Time Collection Calendar Options (3:24)
- Recipe, Half 2: Normalization (Standardization) & Categorical Encoding (5:36)
15.4 Modeling: Make 7 Panel Fashions
- Panel Mannequin 1: Prophet with Regressors (2:11)
- Panel Mannequin 2: XGBoost (2:41)
- Panel Mannequin 3: Prophet Enhance (1:57)
- Panel Mannequin 4: SVM (Radial) (2:02)
- Panel Mannequin 5: Random Forest (1:31)
- Panel Mannequin 6: Neural Internet (1:27)
- Panel Mannequin 7: MARS (1:27)
- Accuracy Examine – It will assist us choose fashions for tuning (3:22)
15.5 Hyperparameter Tuning the Panel Fashions
- Tuning Resamples: Ok-Fold Cross Validation (2:45)
- Panel Mannequin 8: XGBoost Tuned | Tunable Workflow Spec (3:37)
- Panel Mannequin 8: XGBoost Tuned | Hyperparameter Tuning (8:12)
- Panel Mannequin 9: Random Forest Tuned | Tunable Workflow Spec (1:56)
- Panel Mannequin 9: Random Forest Tuned | Hypeparameter Tuning (3:28)
- Panel Mannequin 10: MARS Tuned | Tunable Workflow Spec (2:00)
- Panel Mannequin 10: MARS Tuned | Hyperparameter Tuning (3:07)
15.6 Modeltime Panel Analysis
- Modeltime Desk, Calibration & Accuracy for Panel Information [No Changes] (4:37)
- Forecast Visualization for Panel Information [Use keep_data = TRUE] (4:23)
15.7 Time Collection Cross Validation (Modeltime Resample)
- Time Collection Cross Validation (TSCV) (3:37)
- Modeltime Match Resamples (1:48)
- Modeltime Resample Accuracy (3:53)
- Plot Modeltime Resamples (2:15)
15.8 Ensemble Panel Fashions
- Ensemble Common (Imply) & Sub-Mannequin Choice (2:47)
- Accuracy (Check Set, No Inversion) (1:18)
- Forecast Visualization (Check Set, Inverted) (3:57)
- Accuracy by Group (Check Set, Inverted): summarize_accuracy_metrics() [MAE 46] (4:29)
- Refitted Ensemble & Future Forecast (6:11)
- Ensemble Median: Keep away from Overfitting (3:29)
15.9 Recap
- Congrats – You Simply Forecasted 20 Time Collection Utilizing Panel Information Methods! (2:28)
15.10 Code Checkpoint – Panel Information
- Code Checkpoint [File Download]
[PART 3] Deep Studying with GluonTS
- Welcome to Half 3 – Deep Studying with GluonTS (0:53)
- RStudio IDE Preview Model | Greatest for Working with Python
Module 16 – Setting Up GluonTS & Intro to Reticulate
- What’s a Python Atmosphere? And, why do I would like it?
- Setup [File Download] (1:19)
- R Package deal Set up Necessities (2:30)
16.1 Configuring the “r-gluonts” Python Atmosphere
- GluonTS Atmosphere Setup Overview (2:10)
- Putting in the Python “r-gluonts” Atmosphere (2:15)
- Connecting to the “r-gluonts” Atmosphere (2:48)
- Troubleshooting Set up (2:50)
16.2 Testing Modeltime GluonTS
- Deep Studying Experiment – Predict a Straight Line, Half 1 (3:08)
- Deep Studying Experiment – Predict a Straight Line, Half 2 (3:32)
16.3 Attending to Know Reticulate & Your Python Environments
- Managing Python Environments with Reticulate – Conda & Digital Env (3:18)
- Which Atmosphere am I utilizing & What’s in it? (4:43)
16.4 Utilizing a Customized GluonTS Atmosphere
- Setting Up a Customized Python Atmosphere (6:58)
- Activating (Connecting to) a Customized Python Atmosphere (5:39)
- Reactivating the Default GluonTS Atmosphere (2:13)
16.5 Code Checkpoint – GluonTS Atmosphere Setup
- Code Checkpoint [File Download]
Module 17 – Utilizing GluonTS with Reticulate
- GluonTS Deep Studying | Navigating the Documentation (4:46)
- Setup & Introduction [File Download] (3:27)
- Load Libraries (0:42)
17.1 Reticulated Python Fundamentals
- Reticulated Python, Half 1 (7:00)
- Reticulated Python, Half 2 (4:36)
17.2 Information, Preprocessing, & GluonTS ListDatasets
- Getting the Weekly Transactions Information (1:35)
- Making ready the Full Information for Deep Studying (4:36)
- Making a GluonTS ListDataset from a Information Body (Tibble) (3:10)
- Inspecting a GluonTS ListDataset (5:33)
- Changing from GluonTS ListDataset to Pandas Collection (7:20)
17.3 DeepAR GluonTS Mannequin
- The DeepAREstimator & Coach (8:43)
- Making Our First DeepAR Mannequin (5:14)
- The Prediction (Generator) (3:27)
- Probabilistic Forecasting (5:06)
17.4 Visualizing the Forecast | matplotlib, ggplot, & plotly
- Matplotlib, Half 1 (5:06)
- Matplotlib, Half 2 (3:47)
- ggplot + plotly (Interactive), Half 1 (6:26)
- ggplot + plotly (Interactive), Half 2 (4:43)
17.5 Introducing Modeltime GluonTS
- Modeltime DeepAR | Workflow Advantages (6:56)
- Modeltime DeepAR | Including Extra Epochs (1:17)
17.6 Saving & Loading GluonTS Fashions
- Save & Load | Utilizing GluonTS & Reticulate (6:06)
- Save & Load | Modeltime GluonTS Fashions (3:28)
17.7 Bonus!! GluonTS Deep Issue Fashions
- Making a DeepFactorEstimator (5:11)
- Visualizing the Deep Issue Predictions with Matplotlib (3:17)
17.8 Conclusions & Professional/Cons
- Reticulated GluonTS vs Modeltime GluonTS (Execs & Cons) (4:43)
17.9 Code Checkpoint – GluonTS Reticulate
- Code Checkpoint [File Download]
Module 18 – Time Collection Teams with Modeltime GluonTS
- Deep Studying At Scale (with Modeltime GluonTS)
- Setup [File Download] (2:52)
18.1 Information Assortment & Preliminary Preparation
- Getting the Information | GA Webpage Visits Every day (2:17)
- Full Information | Padding the Information (4:02)
- Different Padding Technique
- Full Information | Log1P Transformation (Goal) (1:01)
- Full Information | Prolong (Future Body) (1:41)
- Full Information | Group-Smart Fourier Collection (2:33)
- Full Information | Group-Smart Including Lagged Options (1:47)
- Full Information | Group-Smart Rolling Options (3:10)
- Full Information | Including a Row ID (0:52)
- Information Ready | skimr::skim() – Be careful for lacking information (2:11)
- Future Information | skimr::skim() – Be careful for lacking information (4:07)
- Break up Information Ready (Prepare/Check) (2:15)
- Visually Examine the Prepare/Check Splits – Examine for lacking teams (3:37)
18.2 Deep Studying Fashions – DeepAR
- Modeltime GluonTS Recipe (4:07)
- DeepAR (Mannequin 1) | Understanding deep_ar() & Coaching Our 1st Mannequin (9:56)
- DeepAR (Mannequin 1) | Mannequin Accuracy Analysis (MAE 0.546) (4:07)
- Ahhh My Mannequin Errored (Skimr to the Rescue!) (3:59)
- DeepAR (Mannequin 2) | Adjusting Hyperparameters (4:19)
- DeepAR (Mannequin 2) | Mannequin Accuracy Analysis (MAE 0.537) (1:49)
- DeepAR (Mannequin 3) | Scaling by Group (3:31)
- DeepAR (Mannequin 3) | Mannequin Accuracy (MAE 0.509) (1:17)
18.3 Deep Studying Fashions – N-BEATS
- N-BEATS (Mannequin 4) | Understanding nbeats() & Coaching Our 1st N-BEATS Mannequin (9:57)
- N-BEATS (Mannequin 5) | Enhancing our mannequin with a brand new loss_function (MAE 0.611) (4:25)
- N-BEATS (Mannequin 6) | Ensemble A number of N-BEATS (7:09)
- N-Beats (Mannequin 6) | Mannequin Accuracy (MAE: 0.544) (3:04)
- Future Forecast | Examine Refitted Fashions (6:01)
18.4 Machine Studying – XGBoost
- Organising the Parallel Processing Backend (1:33)
- Recipes for ML (XGBoost Mannequin) (7:01)
- XGBoost Tunable Mannequin Spec (2:34)
- Hyperparameter Tuning the XGBoost Mannequin (6:20)
18.5 Evaluating Our ML & DL Fashions
- Consider Accuracy on the Testing Set (MAE: 0.527) (4:35)
- Visualize the Testing Set Forecast (2:46)
- Refit & Visualize the Future Forecast (2:40)
18.6 Ensembles of ML & DL Fashions [The Best of Both Worlds]
- Ensembles | Combining ML & DL (MAE: 0.496) (5:54)
- Ensemble | Refitting & Forecasting the Future (4:31)
18.7 DL Wrapup – Saving/Loading Fashions & Conclusions
- Saving | Ensemble & Submodels (5:59)
- Loading | Ensemble & Submodels (4:23)
- Conclusions | Deep Studying with Modeltime & GluonTS (2:40)
- Code Checkpoint [File Download]
CONGRATULATIONS!!! You. Did. It.
- WOO HOO!!! Get YOUR Certificates & a reduction in your subsequent buy! (1:07)
SPECIAL Modeltime Ecosystem: Bonus Classes:
-
In regards to the Particular Bonus Classes
- Hierarchical Forecasting with Modeltime (105:37)
- Modeltime H2O: Forecasting with H2O AutoML (63:43)
- Modeltime Recursive: Autoregressive Forecasting (Lags < Forecast Horizon) | Power Demand (95:16)
- Forecasting Airline Passengers Covid-19 | Modeltime 0.7.0 Updates | PyTorch, GluonTS, World Baselines (93:34)
- The best way to Forecast 100 Time Collection | Modeltime Nested (Iterative) Forecasting (113:01)
–––––
✨about Matt Dancho✨
Founding father of Enterprise Science and common enterprise & finance guru, He has labored with many purchasers from Fortune 500 to high-octane startups! Matt loves educating information scientists on find out how to apply highly effective instruments inside their group to yield ROI. Matt doesn’t relaxation till he will get outcomes (actually, he doesn’t sleep so don’t be suprised if he responds to your electronic mail at 4AM)!
0.0 Average Rating Rated (0 Reviews)