Название: Machine Learning in Transportation: Applications with Examples and CodesАвтор: Niharika Dayyala, Nivedya Madankara Kottayi, Rajib Basu Mallick
Издательство: De Gruyter
Год: 2025
Страниц: 172
Язык: английский
Формат: pdf (true), epub
Размер: 47.7 MB
The book introduces the reader to Machine Learning in transportation. It discusses both simple and advanced concepts including core statistics, data wrangling, data visualization, supervised and unsupervised datamining techniques as well as text mining. The book prepares students to manage data, visualize data and apply appropriate Machine Learning techniques on transportation datasets to derive important insights.
- State-of-the-art techniques in Machine Learning.
- Applications in transportation engineering.
- Includes MATLAB and Python codes.
Machine Learning can be broadly classified into supervised and unsupervised learning.
Supervised learning: In supervised learning, the features or attributes are associated with a target or label. For example, the widths of different types of cracks are associated with the level of cracking in pavements. In this type of learning, predictive models are built that relate a set of input variables to a response variable, which could be discrete (classification model) or continuous (numerical prediction model). In this process, the models are trained to either classify or predict numerical values. Methods include Naive Bayes algorithm, support vector machines (SVMs) and artificial neural networks (ANN).
Unsupervised learning: In this type of learning, models are built to cluster a group of observations on the basis of the structure of the dataset. The K-means unsupervised algorithm is used to group or “cluster” data points around their centroids. The objective is to group or cluster them in such a way that the total variance within the clusters is minimized (objective function). The process is started by assigning random centroids, assigning the data to these centroids, and forming the clusters. The algorithm determines the new centroids for these clusters and then iterates the above process until cluster assignments of the data do not change. In the case of unsupervised learning, the objective is to understand the structure of the dataset and/or to separate the data into different groups based on the similarity of the features. Supervised learning has two primary applications such as regression and classification, whereas unsupervised learning is used for clustering the data.
Visualizing data using graphs is a powerful tool to explore and present data. Data visualization helps identify patterns in data through visual observation. Several data visualizations exist, among which are histograms, bar charts, line charts, pie charts, scatterplots, heatmaps, geospatial maps, etc. Each type of visualization is suitable for specific types of variables. To create data visualizations in R, we use the ggplot2 package. With ggplot2, we can create high-quality visualization in different contexts.
Deep Learning is a subset of Machine Learning, a subset of Artificial Intelligence (AI). In Deep Learning, the applicable concept is that the algorithm extracts the features based on which it can classify or regress, rather than the user having to specify them. The goal of the ANN is to output a function that matches the output, as given by the training data. At each of the layers (parts of the chain structure that makes up an ANN), an approximate function is obtained, and the learning algorithm improves the functions until the final output matches closely with the given information. The term “deep” refers to the depth or the size of the chain structure that is utilized. The application of deep learning mostly uses multiple layers in neural networks that can extract high-level features as part of the learning process for complex relationships between inputs and outputs. The network architecture and the training options should be selected in a way such that the network is progressively improved and is able to make better predictions with new data.
Contents:
Скачать Machine Learning in Transportation: Applications with Examples and Codes