Complete Time Series Data Analysis Bootcamp In R

Explore working with dates in R for time series data. Learn to parse and convert strings to date objects, handle the time component and time zone, and compute date differences.

Install and load the loop data package to standardize dates and times, extract year and month, and create time-based features for nonstandard formats and nyc flights data.

Learn to plot temporal data and convert stock price data into a monthly time series object, then visualize trends with windows to reveal changes from 2001 to 2017.

Explore time series components—trend, seasonality, cyclical, and irregular—and how to decompose monthly dengue data and annual GDP growth to reveal long-term patterns and the presence or absence of seasonality.

Learn to visualize trends in time series data by applying weekly and monthly moving averages, smoothing out outliers and missing values, and comparing actual data to smoothed trends in R.

Explore the trend component in time series data by analyzing Japan's GDP growth from 1985 to 2015, testing for significant negative trends using runs tests and p-values.

Explore methods to identify non-obvious trends in weekly and monthly time series using filters, smoothing, and trend lines, with real data examples like cardiac mortality and Dangi cases.

Visualize monthly variations in time series data and assess seasonality using season plots, month plots, and box plots, then smooth trends with one-year, three-year, and ten-year moving averages in R.

Identify cyclical patterns in time series by applying Fourier transforms to detect dominant frequencies and convert them to time periods, illustrated with monthly temperature data and a China subset.

Learn stl decomposition to separate a time series into trend, seasonal, and remainder components, compare additive and multiplicative seasonality, and enable objective analysis for further time-series work.

Explore identifying additive verses multiplicative seasonality in time series using an order-four moving average, subtracting trend, and estimating average seasonality on quarterly and monthly data.

Assess whether a time series is stationary by examining mean, variance, and autocorrelation with checks and the augmented dickey fuller test; Nottingham data appear stationary, while dengue data do not.

Turn non stationary time series into stationary data through differencing, validate with the ADF test, and assess autocorrelation via plots to demonstrate stationarity using global temperature changes and dengue cases.

Apply log transformation and differencing to monthly tractor sales data to achieve stationarity in mean and variance, removing trend and addressing seasonality.

Seasonal differencing removes seasonal trends in monthly data by subtracting the same month from the previous year, helping achieve stationarity. After ADF testing, apply regular differences as needed.

Detrending time series with linear regression removes the trend to reveal residuals. The lecture uses GDP data, plots GDP versus time, and applies the AIDS test to assess stationarity.

Subtract the yearly mean from GDP data to detrend a 71-year time series when linear regression fails, producing a stationary series confirmed by a subsequent test.

Apply simple exponential smoothing with Holt-Winters to short-term forecasts of CO2 time series with additive seasonality, estimating alpha, beta, gamma and producing 1998–1999 forecasts with confidence intervals.

Explore common forecasting techniques, including mean forecasting, naive forecasting (knife), and seasonal knife forecasting, using beer price data in Australia to compare accuracy against 2007 values.

Learn to forecast saving rates in time series data in R using trailing moving averages, applying one-year, five-year, and ten-year averages and a 12-month rolling mean for next-month predictions.

Explore simple moving averages for forecasting, using a package to apply a 12-month smoothed model on monthly hospital patient counts, with hold-out testing and 95% confidence intervals.

Explore theta lines for forecasting with time series data, including converting daily mortality data to monthly, imputing missing values by interpolation, and generating 12-month forecasts using theta line models.

Learn to forecast on the fly with the profit package in R, generating 365-day future projections for Apple stock, using make_future_data_frame, predict, and visualization.

Apply time series based linear regression to forecast US consumption from income using the forecast package. Compare with ordinary least squares and illustrate scenario forecasting for Apple stock trends.

Learn how to include seasonality and trend in linear regression for forecasting time series, using Melbourne daily temperatures, converting dates, creating time features, aggregating data, and forecasting future values.

Explore how to create lagged and leading time series using shifts by periods, utilizing a built-in object and the stats package to prepare data for time series modeling.

Create weekly lags for the E. coli time series by converting data to a ts object, then to zoo, and generate 1–4 week lags for analysis.

Apply lag regression with an autoregressive model of order one to predict present global temperatures from the previous time period, assess assumptions, and fit a linear model.

See how auto.arima selects parameters for us and forecasts future values using the forecast package in R, with non-seasonal data like U.S. consumption and Apple stock.

Apply automatic ARIMA to a real-life dengue fever time series from the Philippines, including data aggregation, stationarity checks, differencing, model selection, and a 12-month forecast.

Forecast irregular web traffic with ARIMA using four-year terms to reveal seasonality. Train and test split, clean data, and compare Fourier-based models for future values.

Explore bats and tbats models for forecasting time series with multiple seasonalities, decompose data into weekly and monthly components, fit tbats, forecast 12 months, and assess accuracy with backtesting.

Explore linear regression on time series data and why it often underperforms, then learn to use the time decay package to generate a temporal signature for improved forecasting.

Apply linear regression on real life data using the time decay package to forecast bike rental demand, splitting data into training and testing sets and evaluating residuals and forecast accuracy.

Explore selecting best regression model for time series forecasting using H2O AutoML, comparing GBM, random forest, and OLS on beer sales data from 2010–2017, with training, validation, and testing splits.

Forecast time series with gradient boosting in R using the forecast package and past data. Build a model with consumption as the response and income as predictor, then forecast consumption.

Explore artificial neural networks and deep neural networks, from neuron-like inputs to multilayer perceptrons, with backpropagation for supervised learning and activation functions.

Explore neural networks for forecasting time series, using the air passengers data with a training window through 1958, then forecast 1958–1961 with forecast and MLP models.

Learn to forecast humidity over time using recurrent neural networks on temporal weather data, via the Orendain package, training on 2008–2017 data and evaluating with predictions.

Apply recurrent neural networks to time series data, using past four months of chicken prices as predictors. Train with epochs and assess performance on unseen data, noting a 0.58 correlation.

Visualize sudden changes in time series using change point detection, mean-variance assessment, and structural break analysis on Japan's GDP growth from 1985 to 2015.

Detect seasonal and trend breaks in time series using the bfast method, applying log change and relative strength to stock and environmental data with weekly frequencies.

Explore structural change analysis in long-term yearly time series using the globe temp dataset in R, detecting breakpoints and deviations from the norm with regression and confidence intervals.

Explore structural changes in forex regimes by detecting structural breaks in Chinese currency movements using segmented regression and breakpoint analysis of daily exchange data against USD, EUR, and GBP.

Harness GitHub as your data portfolio and use GitHub Desktop to push local code to repositories, create repos, and collaborate via fetch and pull.

Learn to run R inside Google Colab by enabling R in IPython, mounting drive, and executing R code cells to read data and manage packages.

Choose the right database by comparing SQL and NoSQL, and apply a quick rule for SQL data in the complete time series data analysis bootcamp in R.