|
所在平台: Coursera |
课程主页: https://www.coursera.org/learn/feature-engineering-matlab
课程评论：没有评论
课程名称：MATLAB 数据处理与特征工程 课程概述：本课程将基于在《MATLAB 探索性数据分析》中学到的技能，为预测建模奠定基础。该中级课程对需要合并多个来源或时间的数据并对建模感兴趣的任何人都非常有用。适合具有领域知识和一定计算工具接触，但没有编程背景的学习者。要成功完成本课程，学员应具备基本统计学知识（如直方图、均值、标准差、曲线拟合、插值），并已完成《MATLAB 探索性数据分析》课程。 在整个课程中，您将学习如何合并不同数据集中的数据并处理常见场景，例如缺失数据。在课程的最后模块中，您将探索处理文本、音频和图像数据的特殊技术，这些数据在数据科学和更高级的建模中非常常见。课程结束时，您将学会如何可视化数据、清理数据并将其整理为可分析的格式，同时识别必要的特质以回答问题。您将能够可视化数据的分布，并通过目视检查来处理影响准确建模的伪影。 课程大纲： 1. 数据调查：应用探索性数据分析技能于新数据集，探索不同类型的分布，计算偏度和四分位差等数量，并学习多维数据的可视化方法。 2. 数据组织：准备数据进行分析，处理字符串变量以提取关键信息，并将分散在多列中的日期和时间信息整合为一个单一的日期时间变量，效率高地加载和合并多个文件中的数据。 3. 数据清理：清理杂乱数据，找到并解决缺失数据和异常值，使用归一化技术比较具有不同尺度的变量。 4. 发现重要特征：创建新特征以更好地理解数据，评估特征的有效性，以确定其是否对预测有用。 5. 特定领域的特征工程：将前四个模块的概念应用于不同领域，基于基于时间的信号（例如手机加速度计数据）创建和评估特征，使用 MATLAB 中的应用程序进行图像处理并基于分割图像创建特征，使用文本处理技术从非结构化文本中发现特征。
Name:Surveying Your Data
Description:In this module you'll apply the skills gained in Exploratory Data Analysis with MATLAB on a new dataset. You'll explore different types of distributions and calculate quantities like the skewness and interquartile range. You'll also learn about more types of plots for visualizing multi-dimensional data.
Name:Organizing Your Data
Description:In this module you'll learn to prepare data for analysis. Often data is not recorded as required. You'll learn to manipulate string variables to extract key information. You'll create a single datetime variable from date and time information spread across multiple columns in a table. You'll efficiently load and combine data from multiple files to create a final table for analysis.
Name:Cleaning Your Data
Description:In this module you'll clean messy data. Missing data, outliers, and variables with very different scales can obscure trends in the data. You'll find and address missing data and outliers in a data set. You'll compare variables with different scales by normalizing variables.
Name:Finding Features that Matter
Description:In this module you'll create new features to better understand your data. You'll evaluate features to determine if a feature is potentially useful for making predictions.
Name:Domain-Specific Feature Engineering
Description:In this module you'll apply the concepts from Modules 1 through 4 to different domains. You'll create and evaluate features using time-based signals such as accelerometer data from a cell phone. You'll use Apps in MATLAB to perform image processing and create features based on segmented images. You'll also use text processing techniques to find features in unstructured text.
In this course, you will build on the skills learned in Exploratory Data Analysis with MATLAB to lay the foundation required for predictive modeling. This intermediate-level course is useful to anyone who needs to combine data from multiple sources or times and has an interest in modeling. These skills are valuable for those who have domain knowledge and some exposure to computational tools, but no programming background. To be successful in this course, you should have some background in basic statistics (histograms, averages, standard deviation, curve fitting, interpolation) and have completed Exploratory Data Analysis with MATLAB. Throughout the course, you will merge data from different data sets and handle common scenarios, such as missing data. In the last module of the course, you will explore special techniques for handling textual, audio, and image data, which are common in data science and more advanced modeling. By the end of this course, you will learn how to visualize your data, clean it up and arrange it for analysis, and identify the qualities necessary to answer your questions. You will be able to visualize the distribution of your data and use visual inspection to address artifacts that affect accurate modeling.
