This paper proposes an efficient time series matching method using Haar Wavelet Transform for dimensionality reduction and indexing. The authors show that Euclidean distance is preserved in the Haar transformed domain, ensuring no false dismissal. They demonstrate that Haar Wavelet Transform outperforms Discrete Fourier Transform (DFT) in experiments, and propose a new similarity model to handle vertical shifts of time series. A two-phase method is introduced for efficient n-nearest neighbor query in time series databases. The paper discusses the challenges of time series indexing, including dimensionality reduction, completeness, and effectiveness. It compares various transformations such as DFT, Karhunen-Loeve (KL) transform, and Singular Value Decomposition (SVD), highlighting the advantages of Haar Wavelet Transform in preserving Euclidean distance and providing better performance. The Haar Wavelet Transform is shown to have multi-resolution representation and time-frequency localization properties, making it effective for time series indexing. The authors also prove that Euclidean distance is preserved in the Haar transformed domain, ensuring no false dismissal. They introduce a new similarity model to handle vertical shifts and propose an efficient two-phase nearest neighbor query algorithm. Experiments on real stock data and synthetic random walk data show that the Haar Wavelet Transform outperforms DFT in terms of pruning power, page accesses, scalability, and complexity. The method is also shown to be effective for different sequence lengths and database sizes. The paper concludes that the Haar Wavelet Transform provides a more efficient and effective approach for time series indexing and retrieval.This paper proposes an efficient time series matching method using Haar Wavelet Transform for dimensionality reduction and indexing. The authors show that Euclidean distance is preserved in the Haar transformed domain, ensuring no false dismissal. They demonstrate that Haar Wavelet Transform outperforms Discrete Fourier Transform (DFT) in experiments, and propose a new similarity model to handle vertical shifts of time series. A two-phase method is introduced for efficient n-nearest neighbor query in time series databases. The paper discusses the challenges of time series indexing, including dimensionality reduction, completeness, and effectiveness. It compares various transformations such as DFT, Karhunen-Loeve (KL) transform, and Singular Value Decomposition (SVD), highlighting the advantages of Haar Wavelet Transform in preserving Euclidean distance and providing better performance. The Haar Wavelet Transform is shown to have multi-resolution representation and time-frequency localization properties, making it effective for time series indexing. The authors also prove that Euclidean distance is preserved in the Haar transformed domain, ensuring no false dismissal. They introduce a new similarity model to handle vertical shifts and propose an efficient two-phase nearest neighbor query algorithm. Experiments on real stock data and synthetic random walk data show that the Haar Wavelet Transform outperforms DFT in terms of pruning power, page accesses, scalability, and complexity. The method is also shown to be effective for different sequence lengths and database sizes. The paper concludes that the Haar Wavelet Transform provides a more efficient and effective approach for time series indexing and retrieval.