This paper proposes an efficient method for time series matching using the Haar Wavelet Transform (HW). The main contributions include: 1. **Euclidean Distance Preservation**: The HW transform preserves Euclidean distance in the transformed domain, ensuring no false dismissal. 2. **Performance Improvement**: Experiments show that the HW transform outperforms the Discrete Fourier Transform (DFT) in terms of precision and computational complexity. 3. **New Similarity Model**: A new similarity model is introduced to handle vertical shifts in time series, enhancing the effectiveness of the matching process. 4. **Efficient n-Nearest Neighbor Query**: A two-phase method is proposed for efficient n-nearest neighbor queries in time series databases, leveraging the range query capabilities of multi-dimensional index trees like R-Trees. - **Introduction**: Time series data are crucial in various applications, and efficient retrieval methods are essential. The paper discusses the challenges of dimensionality reduction and the limitations of traditional methods like DFT and K-L transform. - **Related Work**: The paper reviews existing techniques, including DFT, K-L transform, and wavelet transforms, highlighting their strengths and weaknesses. - **Proposed Approach**: The HW transform is chosen for its fast computation, low storage requirements, and ability to preserve Euclidean distance. The paper defines a new similarity model that accounts for vertical shifts and proposes an efficient algorithm for n-nearest neighbor queries. - **Performance Evaluation**: Experiments using real stock data and synthetic random walk data demonstrate the effectiveness of the proposed method, showing superior precision, scalability, and performance compared to DFT. - **Conclusion**: The paper concludes by discussing future work, including the potential use of other wavelets and the application of the method to different types of time series data. - **Kin-pong Chan and Ada Wai-chee Fu**: Department of Computer Science and Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong. {kpchan, adafu}@cse.cuhk.edu.hkThis paper proposes an efficient method for time series matching using the Haar Wavelet Transform (HW). The main contributions include: 1. **Euclidean Distance Preservation**: The HW transform preserves Euclidean distance in the transformed domain, ensuring no false dismissal. 2. **Performance Improvement**: Experiments show that the HW transform outperforms the Discrete Fourier Transform (DFT) in terms of precision and computational complexity. 3. **New Similarity Model**: A new similarity model is introduced to handle vertical shifts in time series, enhancing the effectiveness of the matching process. 4. **Efficient n-Nearest Neighbor Query**: A two-phase method is proposed for efficient n-nearest neighbor queries in time series databases, leveraging the range query capabilities of multi-dimensional index trees like R-Trees. - **Introduction**: Time series data are crucial in various applications, and efficient retrieval methods are essential. The paper discusses the challenges of dimensionality reduction and the limitations of traditional methods like DFT and K-L transform. - **Related Work**: The paper reviews existing techniques, including DFT, K-L transform, and wavelet transforms, highlighting their strengths and weaknesses. - **Proposed Approach**: The HW transform is chosen for its fast computation, low storage requirements, and ability to preserve Euclidean distance. The paper defines a new similarity model that accounts for vertical shifts and proposes an efficient algorithm for n-nearest neighbor queries. - **Performance Evaluation**: Experiments using real stock data and synthetic random walk data demonstrate the effectiveness of the proposed method, showing superior precision, scalability, and performance compared to DFT. - **Conclusion**: The paper concludes by discussing future work, including the potential use of other wavelets and the application of the method to different types of time series data. - **Kin-pong Chan and Ada Wai-chee Fu**: Department of Computer Science and Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong. {kpchan, adafu}@cse.cuhk.edu.hk