Python for Time Series Analysis and Forecasting

Master time series fundamentals, including time stamps, ordering, and plotting for seasonality and trend, with autocorrelation, stationarity, ACF/PACF plots, and the augmented Dickey-Fuller test in Python.

Explore the lynx time series dataset, a yearly csv with 114 observations, showing seasonal pulses and autocorrelation, and learn to import it with pandas read_csv.

Explore how time series rely on an ordered index and time stamps, and learn to convert data frames into time series in Python using index and frequency settings.

Read time series plots to recognize trend, seasonality, mean, and variance, and assess stationarity and autocorrelation using visual charts or statistical tests, including the Durbin-Watson test.

Assess stationarity in time series by checking mean, variance, and autocorrelation; apply differencing or transformations to achieve stationarity, then perform the augmented Dickey-Fuller test in Python with a reusable function.

Explore autocorrelation in time series by examining how past observations influence future ones, using lags, Arema models, and visual HCF and PSU plots to detect and interpret serial dependence.

Visualize time series with line plots, using time on the x axis and values on the y axis, as seen in stock charts, and customize with titles and legends.

Learn how moving averages and smoothers tame outliers in time series, using rolling mean and standard deviation, and exponential weighted moving averages with Python and Pandas.

Explore inflation data in python by importing, timestamping, and plotting to reveal trend and seasonality, then assess stationarity with the augmented dickey-fuller test and acf/pacf plots for time series forecasting.

Explore the ARIMA model for time series in Python, from theory of autocorrelation and stationarity to manual parameter selection and residual analysis using the links dataset.

Explore arima models for univariate time series, including p, d, q parameters, differencing for stationarity, and seasonal extensions, with Python examples using the statsmodels tsa module.

Select ARIMA parameters for a univariate series using Arema, test stationarity, and iteratively refine p and q with ACF, PACF, and residual diagnostics; the final model is AR(4,0,0).

Explore how ARIMA residuals reveal model quality in time series analysis by examining mean zero, constant variance, uncorrelated residuals, autocorrelation via plots, and near normal distribution.

Reproduce and explain an ARIMA model by tracing its autoregressive terms, constant, and error using the links dataset, and illustrate in-sample fits and the Kalman filter.

Identify arima parameters using rules for d, p, q, and seasonal components. Use cf, pacf, and hcf plots, test p and q separately, and compare models with information criteria.

Learn how to forecast time series with ARIMA by using predict and forecast functions, compare models in a single plot, and interpret forecasts with a grain of salt.

Analyze the Singapore labor force participation rate from 1980 to 2007 using Gapminder data. Fit ARIMA models and compare p and q with residuals and information criteria.

Analyze a 240 observation temperature dataset from Nottingham (1920–1939) showing clear seasonality with no trend. Learn to import it with read_csv and convert to a monthly time series for modeling.

Explore seasonal decomposition for univariate time series with seasonal components, comparing additive and multiplicative models, and examining strengths and limitations for Python users, with a view toward forecasting.

Subtract the seasonal component from decomposed data to obtain seasonally adjusted series. Use Sdl decomposition with 12-month periods and forecast options like drift, mean, or seasonal naive.

Survey linear and non-linear time series models, including Arema and exponential smoothing, linear regressions, seasonal decomposition, vector autoregressive models, state space models, and neural nets.

Explore exponential smoothing for time series, decomposing data into level, trend, and seasonality with additive or multiplicative options; apply simple exponential smoothing or Holt-Winters in Python.

This lecture demonstrates an additive seasonal exponential smoothing model with no trend for a seasonal dataset with 12 periods. Python automatically estimates alpha and gamma and generates a 12-step forecast.

Explore qualitative forecasting methods for when data is missing, including the Delfi method, forecasting by analogy, and scenario analysis, and learn to integrate them with quantitative methods.

Explore Prophet by Facebook for time series analysis, a decomposition-based model that fits trend, seasonality, and holidays, using regression for each component and enabling change points.

Install the FP profit package, format data with ds and y, fit a monthly seasonality prophet model, and forecast 12 months with yhat and confidence intervals.

Apply seasonal models to the U.S. inflation time series, decompose trend and seasonality, and forecast with exponential smoothing for one full seasonal cycle and one extra year.

Import EU indices data into a pandas data frame with a date time index for a multivariate time series. It includes DAX, SMI, CAC, and FTSE from 1991 to 1998.

Check stationarity in multivariate time series with the augmented dickey-fuller test on each column, drop NAs, difference once and re-test to ensure p-values are below 0.05 for a VAR model.

Explore vector autoregressive models for multivariate time series, where each variable is a linear function of past lags of all variables, with AIC-based order selection and Granger causality testing.

Learn to fit a var model in python and identify the lag order using select_order or direct fit, guided by information criteria to determine the best model.

Apply a multivariate VAR model to forecast three time series, adjust forecasts by inverse differencing to the original scale, and visualize 50-day DAX forecasts with matplotlib.

Explore time series analysis resources, highlighting R as the popular tool, Python interoperability, Hindman’s forecasting principles and practice, and Penn State’s free course.

Import the U.S. inflation data with pandas read_csv and convert it to a monthly time series; extract the inflation rate and assess seasonality with ACF, PACF, and augmented Dickey-Fuller tests.

Analyze the seasonal models homework solution using the U.S. inflation data, apply statsmodels' seasonal decomposition to reveal additive seasonality and no trend, then forecast with a 12-period monthly model.