Time Series Analysis and Forecasting Using Python in 2023

Analyze time series data across daily, weekly, monthly, and quarterly frequencies to identify trends, seasonality, and cyclic patterns. Decompose series into trend, seasonality, and irregular components to forecast future values.

Learn how forecasting uses data to predict future demand, sales, and production, and differentiate it from prediction while considering short-term, medium-term, and long-term forecasts.

Explore additive and multiplicative time series models by decomposing data into trend, seasonality, and noise, and learn how forecasts combine these components for accurate predictions.

Explore population and sample concepts, distinguish a complete population from a random sample with and without replacement, and illustrate how sample data yield mean and standard deviation.

Explore population versus sample concepts, compute sample means, and understand variance and standard deviation (sigma) through practical data points. Learn how these measures describe data spread and central tendency.

Explore forecasting metrics, including simple average, moving averages, and weighted moving averages that emphasize recent data, and contrast neural network based models with seasonality and trend.

Learn how simple average, moving average, and weighted moving average forecast future time steps, with one-step forecasts using the last n values and exponential smoothing techniques.

Install Python through Anaconda across Windows, Mac, and Linux by downloading the installer, launching it, agreeing to terms, and choosing an install location, then open Jupyter Notebook for projects.

Explore essential Python libraries for data science, including NumPy, pandas, Matplotlib, Seaborn, and scikit-learn, and learn when to use them to streamline data manipulation, visualization, and modeling.

Master Python lists by defining them with square brackets, storing mixed data types, accessing and updating elements by zero-based indexing, deleting items, concatenating lists, checking membership, determining length, and iterating.

Sort numeric lists in Python using list.sort to modify the original list or sorted to return a new sorted list, illustrating ascending and descending options.

Explore python lists: how indexing starts at zero, how append vs extend alter list length, and how appending a list creates a list of lists while extending adds elements individually.

Learn how to join a list of strings with a chosen separator in python, recover the original list with split, and compare join with the plus operator as alternatives.

Explore how the Python range function creates sequences in three forms: range(n), range(start, end), and range(start, end, step), with end exclusive, and convert to a list for use in loops.

Explore how the enumerate function pairs list elements with their indices, and harness it in for loops and list comprehensions to create lists, dictionaries, or sets of index-value pairs.

Learn how Python dictionaries map keys to values, create and update entries, access elements by key, remove items with del or pop, and build dictionaries via dict() and comprehension.

Learn how Python sets store unique elements, ignore order, and differ from lists; create sets with set() or curly braces, and convert back to lists when needed.

Show how to write lambda functions as compact alternatives to def, compute x squared, and use map and list comprehension to generate squares from 1 to 5.

Learn how list comprehension in Python creates lists dynamically using for x in range(6), with optional conditions, and extends to dictionaries and nested lists for data analysis.

Open a file, write text with write, and close the file in Python. Then save a dataframe to CSV with to_csv, using a comma delimiter and no index.

Learn how to load time series data, clean and transform it, visualize trends, and apply the Dickey-Fuller test for stationarity in Python.

Explore implementing simple and exponential smoothing time series models in Python, including moving averages, Holt's linear method, and Holt-Winters seasonality, using monthly airline passenger data.

Explore how lags and correlations in time series help predict passengers by shifting data and creating lag features, quantify relationships with correlation, and assess which lag depths yield best predictions.

Explore the arima model for time series forecasting, including autoregressive and moving average components, integration through differencing, and identifying stationarity with lags and errors.

Load and preprocess time series data, convert to a suitable frequency, ensure stationarity through differencing and transformations, then use acf and pacf to select arima p,d,q and forecast future values.

Explore how the Augmented Dickey-Fuller test assesses stationarity in time series data and how differencing and log transforms enable reliable forecasting.

Explore how to plot ACF and PACF in Python to diagnose stationarity, identify AR and MA components, and guide differencing for time series forecasting.

Explore a seasonal model combining integrated moving average with seasonality to forecast monthly sales from daily data by converting to monthly, selecting a category, and fitting the model.

Preprocess the data by filtering to furniture, organizing by category, and computing monthly means. The caption notes seasonality with January dips and June peaks in the time series.

Divide time series data into training and testing sets to validate forecasting models, using the last eleven months as testing data and earlier data for training.

Assess time series stationarity using the Dickey-Fuller test and ACF/PACF plots, apply transformations to stabilize mean and variance, and use lag information to forecast future values.

Decompose time series into trend and seasonal components with an additive model to reveal patterns, and explore SARIMA parameters (p,d,q) with seasonal (P,D,Q) settings for monthly forecasting.

Learn to find the best sarima parameters by evaluating multiple p, d, q and P, D, Q combinations using aic on training data to select the lowest score.

Fit a SARIMA model to time series data, generate forecasts and confidence intervals, and assess diagnostics on residuals to ensure normality and stable variance.

Forecast the next 12 steps using sarima on the 2014–2017 data, compare the predicted mean with actual values, and use confidence intervals to assess forecast accuracy.