Mastering Time Series Forecasting with Python

Sign in with a Gmail account, open Google Colab, create and rename notebooks, organize folders for time series and visualizations, connect the runtime, import libraries, and check Python 3.6.9.

Learn to load data into Google Colab by mounting Google Drive, creating a data folder, uploading Excel files, and loading them into your visualization workflow.

Explore time series visualisations in Python, learning four core chart types—block, area plot, bar plot, and heat maps—for effective time series data visualisations.

Learn to set up Google Colab for time series analysis, mount Google Drive, load Excel and setsu data, and prepare a working directory for visualizing sales data in Python.

load the data by importing numpy, pandas, matplotlib, and seaborn; read an excel file with pandas, inspect data with info and head, and convert sales to datetime64.

Explore line charts for time series by plotting sales over time with Seaborn and Pandas, revealing trend and seasonality and how to adjust size and index for clarity.

visualize time series data with hue in a seaborn line chart to separate series by year and quarter, extract year and quarter from date columns, and apply color maps to reveal patterns.

Learn to create area charts from time series data in Python by setting the index and figure size, and explore stacked area plots, with upcoming preprocessing and visuals coverage.

Group sales by year, compute total sales, and visualize with a bar plot; learn to customize seaborn plots, reset indices, and prepare data for quarterly and stacked bar plots.

Learn to preprocess quarterly sales data, group by year and quarter, reshape into a 9x4 matrix, and create stacked bar and area charts to visualize time series trends.

Explore heatmaps for time series data, using prepared sales data to visualize quarterly trends, customize color maps, axis labels, and value formats to reveal peak sales periods.

Explore simple linear regression, predicting a dependent variable from an independent one with a line y = a + b x, using scatterplots and the least squares approach.

Explore exploratory data analysis to uncover data patterns before modeling. Compute central tendency and spread using median, mode, averages, variance, standard deviation, and Ettakatol range, plus visualization.

Explore quantitative techniques for exploratory data analysis in Python using Google Colab. Load data, compute mean, median, mode, variance, and describe statistics.

Explore exploratory data analysis with graphical techniques in Python, using distribution plots, KDE, histograms, box plots, and scatterplots to understand X and Y distributions, skew, covariance, and outliers.

Develop a simple linear regression by hand, computing covariance of X and Y, variance of X, the slope and intercept, then plot and predict with the regression line.

Learn to build a simple linear regression model in Python with sklearn, fit on X and y, inspect the intercept and coefficients, and predict with a plotted regression line.

Explore building linear regression with statsmodels in Python, defining Y as the dependent and X as the independent variable, fitting the ordinary least squares model, and predicting new values.

Evaluate linear regression by comparing total and explained variance, using sums of squares, mean squared error, and the coefficient of determination (R^2) to judge model fit.

Evaluate a linear regression with mean square error, sum of squares and explained variance to assess model performance. Learn forecasting with regression modeling.

Explore how time as the independent variable drives a linear regression to forecast sales, revealing trends, seasonality, and confidence intervals.

Preprocess time series data for regression forecasting in python by loading sales data in google colab, converting dates to datetime, and creating a time index with month features.

split the time series data into training and testing sets to evaluate a simple linear regression, using time as the independent variable and sales as the dependent variable.

Learn to integrate time into linear regression to model trends in time series data using Python, train the regression model, and evaluate predictions with training and test data.

visualize time series data with a scatterplot and trend line, perform a train-test split, forecast future values with linear regression, and illustrate a 95% confidence interval and margin of error.

Apply the additive model to time series by adding seasonal components to a regression, using N-1 dummy columns for months to capture cyclicality and improve forecasting with confidence intervals.

Explore forecasting with a seasonal additive model in Python, using train data, time and month features, and monthly dummies to capture seasonality, aided by scatterplot visuals.

Create seasonal features by one-hot encoding month names, generate and align dummy variables, and concatenate them with train and test data for consistent time series forecasting.

Split time series data into train and test sets, define time as the independent variable, and prepare corresponding x and y for forecasting models.

Build an additive model using statsmodels by creating and training a linear regression model, importing the linear model, and confirming successful model training.

Apply additive model forecasting and visualizations to generate forecasted values from training and test data, compute the margin of error using standard error, and plot 95% confidence intervals.

Discover multiplicative modeling for time series forecasting by extracting the trend with linear regression, computing seasonal deviation factors, and multiplying forecasts by seasonal effects to predict sales.

Build a multiplicative forecasting model with linear regression on time, derive the trend component, and incorporate seasonal deviation to forecast sales from training and test data.

Calculate the seasonal deviation using the prescribed formula and apply it to the training sales data to forecast values over time, inspecting the top 20 values.

Compute the average seasonal deviation for each month by grouping deviations by month name and averaging across years to form the seasonal corrector factor.

Learn to build multiplicative time series forecasts in Python, computing fitted values, seasonal deviation, and correction factors to forecast future values and assess margin of error.

Calculate the margin of error for the train data using standard error and a 95% confidence interval, where the standard error equals the error and represents the variation of errors.

Visualize forecasted data alongside true values, compute confidence intervals and margins of error, and compare multiplicative and additive models to assess fit over time.

Explore auto regressive methods that forecast future values from past p observations, identify the lag using the partial autocorrelation function, and build AR models for time series forecasting.

Build an autoregressive time series model in Python using the airline passengers data, update statsmodels, and run the workflow in Colab by mounting drive and loading the dataset.

Learn how to preprocess time series data by inspecting data types, converting to datetime, visualizing seasonality, and performing train-test splits to prepare robust forecasts with Python.

Analyze ACF and PACF plots to identify lag order for autoregressive models, detect seasonality, and decide when to remove trend for stationarity.

Learn to make time series stationary by differencing and removing seasonality, then use partial autocorrelation to select lag terms for autoregressive modeling in Python.

Build an auto regressive model for time series forecasting by selecting the lag order p, preprocessing data, and training the model with a single column and month index.

Apply an autoregressive model to generate fitted and forecasted values, compare them to actual data, and build confidence intervals with margins of error.

Evaluate autoregressive models using diagnostic plots, residual normality and randomness tests, and partial autocorrelation to confirm no pattern remains, assess model fit, and forecast with AR models.

Discover three smoothing techniques: naive forecasting, models, and exponential smoothing, for stationary data with no trend, seasonality, or randomness, to generate accurate forecasts.

Explore the naive forecasting model, the simplest time series approach that uses the most recent value as tomorrow's forecast, a single time shift that ignores seasonality.

Apply a naive forecasting model in python by preprocessing airline passenger data in colab, converting month to datetime, visualizing seasonal patterns, and performing a train–test split.

Explore the naive forecasting model in Python, where the next value equals the previous one, using a one-period shift, with forecast versus true values and the margin of error.

Learn the simple moving average as a smoothing method for time series, forecasting future values from past data using an order set by the autocorrelation function.

build a simple moving average model in Python using a rolling window of order three to forecast future values, choosing the window and calculating the mean.

Build a simple moving average model of order q in Python, using a rolling window and mean to forecast future values from the data.

Mastering time series forecasting with Python explains weighted moving average, assigning higher weights to recent values (T-1, T-2, T-3) to improve forecasts and build the model.

Create a three-point weighted moving average model in Python, assigning greater weight to recent observations, computing in-sample training data and out-of-sample forecasts, and contrasting with simple moving averages.

Explore exponential smoothing and exponential moving average, which weight past values with exponentially decreasing importance. Use alpha, the smoothing constant between 0 and 1, to forecast and analyze error.

Build an exponential moving average model in Python to forecast future values from train and test data in pandas, tuning alpha via grid search and comparing to simple moving average.

Explore the ARMA model, which combines autoregressive and moving average components to improve time series forecasts; learn how the AR and MA orders define its structure and parameters.

Explore non seasonal ARIMA, the auto regressive integrated moving average model, and learn how integration and the parameter d determine the non seasonal differences needed to make data stationary.

Learn to build non-seasonal time series models in Python, including ARMA and ARIMA, by loading data in Google Colab and configuring statsmodels.

Split the time series data into training and testing sets, convert the month to a datetime index, and visualize the train vs. test series to prep for arma models.

Convert non-stationary data to stationary, then determine AR order p from the PKF plot and MA order q from the floor to build ARMA models.

Run the Augmented Dickey-Fuller test to determine if data is stationary, compare the p-value to a 5 percent significance level, and convert non-stationary data via differencing with one lag.

Convert non-stationary data to stationary and visualize the stationary series with a simple plot. The p-value 0.00325 leads us to reject the null and prepare the ARMA model.

Explore how to identify ARMA model orders using ACF and PACF on stationary data, then train an ARMA(6,6) model and assess cyclic behavior.

Evaluate your ARMA model by computing the AIC information criterion and mean square error to assess goodness of fit on train data and forecasts, then visualize the model output.

Visualize arma forecast results by plotting train versus predicted values, interpreting the stationary predictions, and preparing to convert forecasts back to non-stationary data in the next lecture.

Convert stationary data to non-stationary by reversing the difference with a cumulative sum, using the last training value as the initial value for arma forecasts with p=6 and q=6.

Build an arima model by selecting p, d, q for stationary data and exploring integration, then visualize forecasts and confidence intervals and convert predictions back to non-stationary.

Visualize ARIMA forecasts by plotting forecasted values with their confidence intervals against the time index. Adjust the lower and upper limits, alpha, and colors to clearly show forecast uncertainty.