Book Image

Python Data Analysis - Second Edition

By : Ivan Idris
Book Image

Python Data Analysis - Second Edition

By: Ivan Idris

Overview of this book

Data analysis techniques generate useful insights from small and large volumes of data. Python, with its strong set of libraries, has become a popular platform to conduct various data analysis and predictive modeling tasks. With this book, you will learn how to process and manipulate data with Python for complex analysis and modeling. We learn data manipulations such as aggregating, concatenating, appending, cleaning, and handling missing values, with NumPy and Pandas. The book covers how to store and retrieve data from various data sources such as SQL and NoSQL, CSV fies, and HDF5. We learn how to visualize data using visualization libraries, along with advanced topics such as signal processing, time series, textual data analysis, machine learning, and social media analysis. The book covers a plethora of Python modules, such as matplotlib, statsmodels, scikit-learn, and NLTK. It also covers using Python with external environments such as R, Fortran, C/C++, and Boost libraries.
Table of Contents (22 chapters)
Python Data Analysis - Second Edition
Credits
About the Author
About the Reviewers
www.PacktPub.com
Customer Feedback
Preface
Key Concepts
Online Resources

Filtering


Filtering is a type of signal processing, which involves removing or suppressing a part of the signal. After applying FFT, we can filter high or low frequencies, or we can try to remove the white noise. White noise is a random signal with a constant power spectrum and as such doesn't contain any useful information. The scipy.signal package has a number of utilities for filtering. In this example, we will demonstrate a small sample of these routines:

  • The median filter calculates the median in a rolling window (see http://en.wikipedia.org/wiki/Median_filter). It's implemented by the medfilt() function, which has an optional window size parameter.

  • The Wiener filter removes noise using statistics (see http://en.wikipedia.org/wiki/Wiener_filter). For a filter g(t) and signal s(t), the output is calculated with the convolution (g * [s + n])(t). It's implemented by the wiener() function. This function also has an optional window size parameter.

  • The detrend filter removes a trend. This can...