| Author: | Chui, Wai Hin |
| Title: | Multivariate time series prediction and classification using machine learning |
| Advisors: | Chan, Keith (COMP) |
| Degree: | M.Sc. |
| Year: | 2017 |
| Subject: | Hong Kong Polytechnic University -- Dissertations Machine learning Computer algorithms |
| Department: | Department of Computing |
| Pages: | viii, 47 pages : color illustrations |
| Language: | English |
| Abstract: | As the technology advances over the years, there are many factors favouring the growth of data. The drop in price of storage device for example, allows users not to worry about data size and record everything. This comes to another factor that is the advancement in measurement or recording devices and algorithms. Many different things can now be recorded in high precision and effciency like ask bid price from stock exchanges, sensors data collected from industrial plants, user behaviour from internet services like web surfng history and social media activity. Not only can these data be stored easily, they can be transmitted and replicated with little efforts. Due to fact that many different type of data can be collected about a single object or event, the diversity and dimension of data are now very large. If one can mine meaningful patterns from this high volume data, the patterns can be used in forecasting the future and better prepare ourselves to reduce risk ahead of time. Financial industry for instance, institution can predict stock price movement by integrating different sequences such as index movements, price movements of other stocks, probably the sentiment movement and corporate announcement [2] about a particular listed company into the forecasting model so that they can access the risk ahead of time and take action earlier to hedge or eliminate the risk. Besides fnancial market, the electric power supply agencies can make use of time series forecasting to model the yield and load of their power plants which could be every important particularly for renewable energy source like wind and solar power that depend greatly on the weather and other environmental factors. The model can also model the varying electric power demand of their customers and help them maximize the use of renewable energy instead of burning fossil fuel [1]. The model can well be extended to monitor also the exhaust emission of fossil fuel based power plants to control the pollution. Therefore, a interpretable, effective and scalable algorithms that can handle large multivariate time series data are very valuable especially in the big data era which is also the motivation of this paper. If there exist a less restrictive scalable approach that could extract important temporal patterns effciently from different sequences, even less experienced persons would be able to make use of data and have more intuitions about the problem allowing them to make better informed decisions. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 991021953150903411.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 688.57 kB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/9010