Times Series Analysis for Everyone LiveLessons

Times Series Analysis for Everyone LiveLessons

English | MP4 | AVC 1280×720 | AAC 44KHz 2ch | 6h 02m | 1.49 GB

Times Series Analysis for Everyone LiveLessons covers the fundamental ideas and techniques for the analysis of time series data. This course introduces you to the basic concepts, ideas, and algorithms necessary to develop your own time series applications in a step-by-step and intuitive fashion. The lessons follow a gradual progression, from the more specific to the more abstract, taking you from the very basics to some of the most recent and sophisticated algorithms.

Learn How To

  • Use Pandas for time series
  • Create visualizations of time series
  • Transform time series data
  • Apply Fourier analysis
  • Utilize time series correlations
  • Understand random walk models
  • Explore and fit ARIMA models
  • Explore and fit ARCH models
  • Integrate machine learning into time series analysis
  • Integrate deep learning into time series analysis

Lesson 1: Pandas for Time Series
Pandas was originally developed for financial applications. As such, it was developed with time series support from day one. In this lesson we review some of the fundamental features of pandas that we use in the remainder of the course.

Lesson 2: Visualizing Time Series Modeling
Visualization is a fundamental first step when exploring and understanding a new dataset. Here we visualize and highlight important features of the example time series we will analyze in detail.

Lesson 3: Stationarity and Trending Behavior
Time series can exhibit characteristic types of behavior, such as trends, seasonal, and cyclical patterns. In this lesson you learn how to identify each of these behaviors and to remove them from the time series in order to facilitate its analysis.

Lesson 4: Transforming Time Series Data
The modeling and analysis of time series often require us to transform the original data. In this lesson we learn how to calculate and apply the most common transformations, how to impute missing data, and how to estimate basic properties of the time series.

Lesson 5: Running Value Measures
Perhaps the simplest time series analysis you can perform is the exploration of how various metrics evolve as a function of time. In this lesson you learn how to calculate measures using running windows.

Lesson 6: Fourier Analysis
Fourier analysis is a powerful tool. In this lesson we explore how it enables us to not only observe the strongest frequencies present in the data, but also to eliminate noise patterns and perform simple extrapolations of future values.

Lesson 7: Time Series Correlations
An important step in characterizing a time series is understanding how it correlates with itself. The auto-correlation and partial-auto-correlation functions are the two most important functions we use to determine the temporal properties of our time series.

Lesson 8: Random Walks
A random walk—a sequence of positions where each step is chosen at random—is perhaps the simplest example of time series. Here we use it as a prototypical model to understand the fundamental ideas behind time series analysis and to determine whether or not a given series is stationary.

Lesson 9: ARIMA Models
The ARIMA class of models is the most popular and well-known family of time series models. It relies on the concepts of partial and full auto-correlation to define a simple random walk-like process that is able to reproduce most time series in a simple and efficient manner.

Lesson 10: ARCH Models
The ARIMA class of models requires the underlying time series to be stationary. When that assumption is not true, we need to rely instead on the ARCH class of models that generalizes ARIMA to the situation, common in financial time series, in which the variance of the time series changes over time.

Lesson 11: Machine Learning with Time Series
Both ARIMA and ARCH models are classical models that were developed specifically for the modeling of time series. However, it is possible to apply a wide range of machine learning approaches to the modeling and forecasting of time varying phenomena.

Lesson 12: Overview of Deep Learning Approaches
Recurrent neural networks are a class of deep learning architectures that were developed specifically to be used in modeling sequential patterns such as sequences of words, sounds, and other related phenomena. In this lesson you learn how you can apply them directly to time series.

Table of Contents


Lesson 01 Pandas for Time Series
Learning objectives
1.1 DataFrames and Series
1.2 Subsetting
1.3 Time Series
1.4 DataFrame Manipulations
1.5 Pivot Tables
1.6 Merge and Join
1.7 Demo Number

Lesson 02 Visualizing Time Series
Learning objectives
2.1 Data Representation
2.2 Gross Domestic Product
2.3 Influenza Mortality
2.4 Sun Activity
2.5 Dow Jones Industrial Average
2.6 Airline Passengers
2.7 Demo

Lesson 03 Stationarity and Trending Behavior
Learning objectives
3.1 Non-stationarity
3.2 Trend
3.3 Demo Number 1
3.4 Seasonality
3.5 Time Series Decomposition
3.6 Demo Number 2

Lesson 04 Transforming Time Series Data
Learning objectives
4.1 Lagged Values
4.2 Differences
4.3 Data Imputation
4.4 Resampling
4.5 Jackknife Estimators
4.6 Bootstrapping
4.7 Demo

Lesson 05 Running Value Measures
Learning objectives
5.1 Windowing
5.2 Running Values
5.3 Bollinger Bands
5.4 Exponential Running Averages
5.5 Forecasting
5.6 Demo

Lesson 06 Fourier Analysis
Learning objectives
6.1 Frequency Domain
6.2 Discrete Fourier Transform
6.3 FFT for Filtering
6.4 Forecasting
6.5 Demo

Lesson 07 Time Series Correlations
Learning objectives
7.1 Pearson Correlation
7.2 Correlation of Two Time Series
7.3 Auto-Correlation
7.4 Partial Auto-Correlation
7.5 Demo

Lesson 08 Random Walks
Learning objectives
8.1 What Is a Random Walk
8.2 White Noise
8.3 Stationary versus Non-Stationary
8.4 Dicky-Fuller Test
8.5 Hurst Exponent
8.6 Demo

Lesson 09 ARIMA Models
Learning objectives
9.1 Moving Average (MA) Models
9.2 Autoregressive (AR) Model
9.3 ARIMA Model
9.4 Fitting ARIMA Models
9.5 Statsmodels for ARIMA Models
9.6 Seasonal ARIMA
9.7 Demo

Lesson 10 ARCH Models
Learning objectives
10.1 Heteroscedasticity
10.2 Hertoscedastical Models
10.3 Autoregressive Conditionally Heteroscedastic (ARCH) Mod
10.4 Fitting ARCH models
10.5 Demo

Lesson 11 Machine Learning with Time Series
Learning objectives
11.1 Interpolation
11.2 Types of Machine Learning
11.3 Regression and Classification
11.4 Cross-validation
11.5 Caveats When Working with Time Series
11.6 Demo

Lesson 12 Overview of Deep Learning Approaches
Learning objectives
12.1 Feed Forward Networks (FFN)
12.2 Recurrent Neural Networks (RNN)
12.3 Gated Recurrent Units (GRU)
12.4 Long Short-term Memory (LSTM)
12.5 Demo