Forecasting Models & Time Series Analysis for Business in R

Explore time series data used in forecasting, illustrated by NYC daily shelter demand from 2014–2020. Recognize time-ordered, equally spaced observations (daily, weekly, monthly, yearly) and patterns in one-dimensional time data.

Learn how to install and load libraries in R, set a working directory, and navigate the RStudio interface for scripts, console, plots, and data frames.

Diogo explains learning to program in R and Python through project-based, goal-driven practice, starting with Stata in business studies, replicating R techniques in Python, and seeking colleague support.

Prepare and run a seasonal decomposition workflow in R using the forecast package, building on a prior script, setting the working directory, loading data, and examining multiplicative seasonality.

Learn to transform data into time series objects in R using RTS, set start and frequency, handle daily and seasonal cycles, and plot the results.

Explore how to create a seasonality plot in R with the DG Season Plot function, mapping X and Y to reveal monthly patterns and potential COVID impacts.

Explore multiplicative decomposition of a time series in R to separate trend, seasonality, and error, and see how removing them clarifies the forecast error as the key.

Explore how error modelling and regression help you capture trend and seasonality in stock data, considering events, temperature, and other drivers, while judging data representativeness.

Provide feedback to shape the remaining half of the course on forecasting models and time series analysis for business in R and share what's exciting you and what's missing.

Explore a case study using google trends data to build, train, test, and visualize a time series in R, applying Holt Winters triple exponential smoothing to forecast 10 weeks.

Learn to load libraries and data in R, set a working directory, read a CSV, and prepare a time series object for exponential smoothing and cold winters forecasting.

Transform a data frame into a time series object in R using the time series function, setting start and a weekly frequency of 52 for weekly data and season plots.

Explore time series plotting in R with plots and seasonal plots, revealing year-end peaks around Christmas in Brazil and a 2020 pandemic drop, plus training and test sets for forecasting.

Forecast air miles with Holt-Winters in R by visualizing data, splitting training and test sets (last 12 months), forecasting 12 months, and measuring accuracy with the TSA package.

Apply Holt-Winters forecasting in R to air miles time series, visualize trends and seasonality, and compare additive and multiplicative models using train-test splits and accuracy metrics.

Develop stakeholder management by involving them early, ensuring they are heard, and sharing updates on what you do that concerns them, as shown in a fashion e-commerce project.

Motivate the building of a reusable forecasting product by ensembleing three models to address model selection uncertainty and reduce flaws, and showcase parameter tuning for Facebook profit and neural networks.

Explore arima forecasting in r, focusing on auto regressive, integrated, and moving average components, seasonality, and exogenous regressors for accurate forecasts.

Learn to prepare an R forecasting script: set the working directory, load forecast libraries, transform dates, select time series and regressors, and inspect seasonal patterns for accurate forecasts.

Learn to split data into training and test sets for forecasting 31 days in R, using a filter-based approach to define date-based training up to a cutoff and test afterward.

Learn to create a time series object in R using the ts function, specifying data and a frequency to capture seasonality, such as weekly or monthly patterns.

Explore Pearson correlation as a metric for strength and direction between variables in data. Values range from -1 to 1, indicating positive or negative relationships, and causation is not implied.

Explore auto correlation plots to assess how past values inform current observations in time series, using lagged correlations that decay toward zero.

Generate and interpret auto-correlation plots in R with the acf function to assess past dependence. See strong autocorrelation near one and decay over lag to gauge time series predictability.

Set the frequency to 365 in the R-based forecasting workflow, inspect autocorrelation, and determine that seasonality is not detected, yielding a non seasonal model with similar predictions.

Explore the SARIMAX forecasting challenge in R by loading a 261-observation time series, creating training and test sets, and forecasting 10 weeks with a SARIMAX model, visualizing and assessing results.

Explore sarimax forecasting in R by building training and test splits, incorporating exogenous variables like Easter, Carnival, and Republic, and evaluating forecast accuracy for a ten-step ahead horizon.

Apply a product-minded forecasting approach: interview stakeholders, define success, plan backward, map external events and controllable factors, ensure exogenous regressions are significant, and iterate to drive demand.

Identify how structural time series decomposes data into trend, seasonality, exogenous impact, and error, and learn to balance explanation with the risk of overfitting.

This lecture shows how to set up the workspace, load Facebook profit and forecast, prepare a 16-variable data set, rename the date column, and create and combine three forecasts.

Create training and test sets in R by filtering the time series to December 2020, excluding Easter and Thanksgiving, and selecting Y, Christmas, and temperature for a Facebook profit model.

Explore the Facebook Prophet model parameters, including holidays, dates, lower and upper windows, seasonality modes (multiplicative or additive), and prior scales to tune trend flexibility.

Review regressor coefficients in a multiplicative forecasting model and read them as percent; temperature lowers demand per degree, while Christmas has a tiny effect, and events will be analyzed next.

Learn to create a future data frame with daily periods using make_future_data_frame, include regressors, then forecast with predict, and analyze holiday and seasonal effects in visualization.

Explore how to inspect event impacts in R using a forecast data frame. Filter for Easter and Thanksgiving dates, and interpret absolute values and negative Easter effects.

Forecast daily Udemy Wikipedia visits using Facebook Prophet in R, incorporating Easter and Christmas as events and Black Friday, then visualize results and assess 31-day forecast accuracy.

Explore how to prepare time series data in R for Facebook Prophet, including date formatting, holiday creation (Easter, Christmas), and training versus test set construction.

Build a Facebook Prophet model in R with yearly and weekly seasonality, holy days, and a Black Friday regressor; forecast future demand and evaluate accuracy with MAE, RMSE, and MAPE.

Tune prophet parameters to optimize forecasts, moving beyond default values with a careful, code-by-code approach. Follow slow, hands-on practice, review questions in the Q&A, and repeat steps in the challenge.

Explore parameter tuning for forecasting models, beginner friendly yet advanced, by defining parameter options, testing adaptive and multiplicative seasonality, and selecting the lowest-error settings.

Explore cross-validation to boost forecasting credibility by iterating multiple training and test splits, moving the test set into training and testing across different seasonality, such as December and June.

Prepare an R script for a Facebook Prophet forecasting model, tune parameters, clean the template, merge holidays, and build a two-regressor model ready for cross-validation.

Apply time series cross-validation in R to set horizon, units, and period, then measure model accuracy with MAPE, MAE, and RMSE.

Explore how parameter tuning tailors forecasting models for accuracy by adjusting seasonality, prior scale, holiday prior scale, change point scale, and seasonality mode (multiplicative or additive).

Create a parameter grid and results vector to tune a Prophet-based forecast. Generate all combinations and perform 24 cross-validations to find best settings.

Learn to tune forecasting models in R by building a parameter grid, configuring seasonality and regressors, and validating results with cross validation to improve accuracy.

Practice parameter tuning by using a grid to link results, identify the minimum, and select best additive seasonality, change points prior scale, and holidays prior scale.

Begin the challenge for forecasting models and time series analysis in R by tuning parameters through cross validation, using a prepared template, with no dataset provided for now.

Demonstrates parameter tuning for time series forecasting in R using Prophet, building a grid, running cross-validation, and selecting best change point and seasonality priors.

Uber employs spatial-temporal and marketplace forecasting to predict demand, supply, revenue, goods to be ordered, and staffing, while comparing statistical and machine learning methods for robust, data-driven decisions.

Learn how neural networks model time series with autoregression by transforming inputs through a hidden layer of neurons into outputs using sigmoid activation and addressing overfitting.

set the working directory, load the forecast library in R, transform the data, rename the variable, and visualize to prepare for forecasting models and time series.

Prepare training and test sets, transform the time series, and extract regressors in R for forecasting, with automated frequency options and neural networks for regression.

Predict employment rate with a neural networks model by splitting the dataset into training and four-period test sets, building a regressive matrix, and using fpp3 in R to assess accuracy.

Explore neural networks for forecasting in R by building test sets, transforming data, and fitting a forecast model; evaluate accuracy with mean absolute error and root mean square error.

Explore parameter tuning of a deep learning time series model, adding a cross-validation loop to the existing tuning process, with emphasis on debugging via code templates and managing programming difficulty.

Explore tuning parameters: p, P, size, and decay for time series forecasting in neural networks, considering non-seasonal and seasonal lags, inputs, and external regressors.

Build a custom cross-validation pipeline and parameter tuning for time series forecasting in R. Prepare data, train-test split, and use neural networks and autoregression to measure accuracy.

Generate a cross-validation date sequence to split training and test sets for forecasting, using 31-day test windows and a 200-day lookback, and implement a repeatable date-based validation pipeline.

Create a results vector to store model outputs, using vector() with mode numeric and length equal to the number of dates (25 forecasts) for cross-validation.

Implement a practical cross-validation pipeline in R for forecasting time series. Build training and test sets, create time series objects with regressors, forecast, and assess RMSE accuracy.

Prepare parameter tuning by building a comprehensive search grid in R, using expand.grid for four parameters, and setting up cross-validation for neural networks to compare results.

Tune forecasting models in R by performing grid search, fetching parameters, and evaluating with cross-validation, then store and summarize the mean results.

Fetch the best parameters in R via a grid search, inspect the results grid for the minimum, and identify decay and size for seasonal forecasts.

Learn to perform cross-validation and hyperparameter tuning on an employment rate prediction case study in R, using a focused script and the Neural Networks Challenge template.

Explore time series forecasting with cross validation and grid search in R, tuning AR and seasonal parameters to optimize mean absolute error and identify best parameter settings.