한국해양대학교

목차

Chapter 1 INTRODUCTION 1
1.1. Research Objectives and Necessities 6
1.2. Research Environment and Methodology 9
1.2.1. Research Environment 9
1.2.2. Research Methodology 10
Chapter 2 RELATED RESEARCH 13
2.1. Data Mining and Machine Learning 13
2.1.1. Data Mining 13
2.1.2. Association Rules Exploration 14
2.1.3. Machine Learning and Application in Data Mining 16
2.2. Time Series Theory 18
2.2.1. Time Series 18
2.2.2. Time Series Forecasting Models 20
2.2.3. Hybrid for Time Series Models 22
2.3. Rough Set Theory 24
2.3.1. Equivalence relation: equivalence class 25
2.3.2. Reduced attributes 26
2.3.3. Differential matrix and distinction function 27
2.4. Consumer behavior and Housing Price Prediction Research 29
2.4.1. Consumer behavior Prediction 29
2.4.2. Housing Price Prediction Research 32
2.5. Technology Platforms for Data Mining and Machine Learning 35
2.5.1. Object Relational Mapping (ORM) 35
2.5.2. Big Data Processing with Apache Spark 37
2.5.3. ML.NET Model Machine Learning Framework 40
2.6. Principles for Evaluation Metrics 43
2.6.1. Mean Absolute Error (MAE) 43
2.6.2. Mean Squared Error (MSE) 44
2.6.3. Root Mean Square Error (RMSE) 44
2.6.4. R-squared 45
Chapter 3 BUILDING NEW MACHINE LEARNING MODEL FOR FORECASTING 46
3.1. Research Motivation for building a new ML.RealEstate 46
3.2. Proposal for New ML.RealEstate Model 48
3.3. Database and Classes Architecture of ML.RealEstate Model 50
3.3.1. Database Structure for House Sale Transaction 50
3.3.2. Classes Architecture of ML.RealEstate Model 51
3.3.3. ML.RealEstate.Data namespace 52
3.3.4. ML.RealEstate.Error Namespace 54
3.3.5. ML.RealEstate.Predict Namespace 55
3.4. BrokerRealEsate Executor for ML.RealEstate Model 56
3.4.1. ImportDataset() Algorithm 58
3.4.2. BuildModel() algorithm 59
3.4.3. Evaluate() algorithm 60
3.4.4. SaveModel() algorithm 61
3.4.5. LoadModel() algorithm 62
3.4.6. Predict() algorithm 63
3.5. Experiment and Results 63
3.5.1. Preparing Dataset 64
3.5.2. Reference the DLL and Run the Model 67
3.5.3. Implementation of ML.RealEstate Model 67
3.5.4. Performance and Evaluation Metrics 73
3.5.4.1. Summary for Evaluation Metrics 73
3.5.4.2. R-Squared with Training Set Ratio 74
3.5.4.3. Comparison of MAE and RMSE Metric 75
3.5.4.4. Comparison of Prediction Home Price 76
3.6. Publish ML.RealEstate Model into NuGet Ecosystem 77
Chapter 4 IMPROVING FORECASTING WITH NEW HYBRID MACHINE LEARNING MODEL 79
4.1. Motivation for new Lucy Hybrid model 79
4.2. Proposal for new Lucy Hybrid model 80
4.3. Classes Architecture and Algorithms of Lucy Hybrid Model 82
4.3.1. Classes Architecture 83
4.3.2. LucyDataExecutor Class 84
4.3.3. LucyMetric Class 87
4.3.4. LucyUtil Class 87
4.3.5. LucyHybrid Class 89
4.3.6. LucyTrendForecast Class 92
4.4. Experiment and Results 94
4.4.1. Preparing Dataset 94
4.4.2. Data Preprocessing 96
4.4.3. Prediction and Evaluation Model 97
4.4.4. Persistent and Comparison of Lucy Hybrid Models 100
4.4.5. Forecasting Experiments 104
4.5. Forecasting Comparison 107
4.6. Reuse Lucy Hybrid Model 111
Chapter 5 BUILDING NEW MACHINE LEARNING MODEL FOR DECISION MAKING 117
5.1. Research Motivation for Building Model with Rough Set 117
5.2. Proposal for Model with Rough Set 118
5.3. Model class and Algorithm Analysis 120
5.3.1. Overall Object Classes for Model 121
5.3.2. Class model for RoughSet 125
5.3.3. Class Model for RoughSetHistory 128
5.3.4. RoughSetUtility Class 130
5.3.5. Main Algorithms Analysis 132
5.3.5.1. Filtering Big Data Using Apache Spark for .NET 132
5.3.5.2. Distinguishing Matrix Finder Algorithm 133
5.3.5.3. Discriminant function algorithm 134
5.3.5.4. Find Reduced Function algorithm 135
5.3.5.5. Decision Rules Algorithm 136
5.4. Experiment and Results 138
5.4.1. Overview the Rough Sets Executor System 138
5.4.2. Rough Set Database Screen 140
5.4.3. Attribute format for Rough Set Database 142
5.4.4. Download machine learning database function 143
5.4.5. Main System of Rough Sets Executor 145
5.4.6. Export to Website function 153
5.4.7. Printing for rough set processing 155
5.4.8. Performance and summary metric with Stages for All jobs 156
Chapter 6 CONCLUSION AND FUTURE WORK 158
Reference 162
Acknowledgements 176