Applications of Machine Learning and Data Analytics Models in Maritime Transportation

Contents
About the Authors
1 Introduction of maritime transportation
1.1 Overview of maritime transport
1.2 World fleet structure
1.2.1 Bulk carrier
1.2.2 Oil tanker
1.2.3 Container ship
1.3 Key roles in the shipping industry
1.3.1 Ship owner
1.3.2 Ship operator
1.3.3 Ship management company
1.3.4 Flag state
1.3.5 Classification society
1.3.6 Charterer
1.3.7 Freight forwarder
1.3.8 Ship broker
1.4 Container liner shipping
References
2 Ship inspection by port state control
2.1 Key issues in maritime transport
2.1.1 Maritime safety management
2.1.2 Marine pollution control
2.1.3 Seafarers' management
2.2 Port state control
2.2.1 The background and development of PSC
2.2.2 Ship selection in PSC
2.2.3 Onboard inspection procedure
2.2.4 Inspection results
2.3 Data set used in this book
References
3 Introduction to data-driven models
3.1 Predictive problem and its application in maritime transport
3.1.1 Introduction of predictive problem
3.1.2 Examples of predictive problem in maritime transport
3.1.3 Comparison of theory-based modeling and data-driven modeling
3.1.4 Popular data-driven models
3.1.4.1 Statistical modeling
3.1.4.2 ML modeling
References
4 Key elements of data-driven models
4.1 Comparison of three popular data-driven models
4.2 Procedure of developing ML models to address maritime transport problems
4.2.1 Problem specification
4.2.2 Feasibility assessment
4.2.3 Data collection
4.2.3.1 Automatic identification system
4.2.3.2 Ship specifications
4.2.3.3 Ship sailing records
4.2.3.4 Ship accident data
4.2.3.5 Port statistics
4.2.4 Feature engineering
4.2.4.1 Data cleaning
4.2.4.2 Feature extraction
4.2.4.3 Feature pre-processing
4.2.4.4 Feature encoding and scaling
4.2.4.5 Feature selection
4.2.4.6 Pearson’s correlation coefficient (Pearson’s ‍r )
4.2.4.7 Spearman’s rank coefficient (Spearman’s ρ)
4.2.4.8 Chi-squared test ( χ2 test)
4.2.4.9 Mutual information
4.2.4.10 Recursive feature elimination
4.2.4.11 Feature selection based on regularization
4.2.4.12 Feature selection based on feature importance
4.2.5 Model construction
4.2.5.1 Model evaluation metrics
4.2.5.1.1 Accuracy
4.2.5.1.2 Precision
4.2.5.1.3 Recall
4.2.5.1.4 F1 score
4.2.5.1.5 Receiver operating characteristic (ROC) curve
4.2.5.1.6 Area under ROC curve (AUC)
4.2.5.2 Model selection
4.2.6 Model refinement
4.2.7 Model assessment, interpretation/explanation, and conclusion
References
5 Linear regression models
5.1 Simple linear regression and the least squares
5.2 Multiple linear regression
5.3 Extensions of multiple linear regression
5.3.1 Polynomial regression
5.3.2 Logistic regression
5.4 Shrinkage linear regression models
5.4.1 Ridge regression
5.4.2 LASSO regression
References
6 Bayesian networks
6.1 Naive Bayes classifier
6.2 Semi-naive Bayes classifiers
6.3 BN classifiers
References
7 Support vector machine
7.1 Hard margin SVM
7.2 Soft margin SVM
7.3 Kernel trick
7.4 Support vector regression
Reference
8 Artificial neural network
8.1 The structure and basic concepts of an ANN
8.1.1 Training of an ANN model
8.1.2 Hyperparameters in an ANN model
8.2 Brief introduction of deep learning models
References
9 Tree-based models
9.1 Basic concepts of a decision tree
9.2 Node splitting in classification trees
9.2.1 Iterative dichotomizer 3 (ID3)
9.2.2 C4.5
9.2.3 Classification and regression tree (CART)
9.2.4 Node splitting in regression trees
9.3 Ensemble learning on tree-based models
9.3.1 Bagging
9.3.2 Boosting
9.3.2.1 Adaboost
9.3.2.2 Boosting tree
References
10 Association rule learning
10.1 Large item sets
10.2 Apriori algorithm
10.3 FP-growth algorithm
References
11 Cluster analysis
11.1 Distance measure in clustering
11.1.1 Distance measure of examples
11.1.2 Distance measure of clusters
11.2 Metrics for clustering algorithm performance evaluation
11.2.1 Clustering algorithms
11.2.2 K-means (partition-based method)
11.2.3 DBSCAN (density-based method)
11.2.4 Agglomerative algorithm (hierarchy-based methods)
Reference
12 Classic and emerging approaches to solving practical problems in maritime transport
12.1 Topics in maritime transport research
12.2 Research methods and their specific applications to maritime transport research
12.3 Issues of adopting data-driven models to address problems in maritime transportation
12.3.1 Data
12.3.2 Model
12.3.3 User
12.3.4 Target
References
13 Incorporating shipping domain knowledge into data-driven models
13.1 Considering feature monotonicity in ship risk prediction
13.1.1 Introduction of monotonicity in the ship risk prediction problem
13.1.2 Integration of monotonic constraint into XGBoost
13.2 Integration of convex and monotonic constraints into ANN (artifical neural network)
References
14 Explanation of black-box ML models in maritime transport
14.1 Necessity of black-box ML model explanation in the maritime industry
14.1.1 What is the explanation for ML models
14.1.2 Propose and evaluate explanations for black-box ML models
14.2 Popular methods for black-box ML model explanation
14.2.1 Forms and types of explanations
14.2.2 Introduction of intrinsic explanation model using DT as an example
14.2.3 SHAP method
References
15 Linear optimization
15.1 Basics
15.2 Classification of linear optimization models according to solutions
15.3 Equivalence between different formulations
15.4 Graphical method for models with two variables
15.5 Using software to solve linear optimization models
15.6 An in-depth understanding of linear optimization
15.7 Useful applications of linear optimization solvers
16 Advanced linear optimization
16.1 Network flow optimization
16.2 Dummy nodes and links
16.3 Using linear formulations for nonlinear problems
16.4 Practice
References
17 Integer optimization
17.1 Formulation I: natural integer decision variables
17.2 Formulation II: 0-1 decision variables
17.3 Formulation III: complex logical constraints
17.4 Solving mixed-integer optimization models
17.5 Formulation IV: challenging problems
17.6 Formulation V: linearizing binary variables multiplied by another variable
17.7 Practice
Reference
18 Conclusion
18.1 Summary of this book
18.2 Future research agenda
Index