Top PDF Wavelet discrete transform, ANFIS and linear regression for short-term time series prediction of air temperature

Wavelet discrete transform, ANFIS and linear regression for short-term time series prediction of air temperature

Wavelet discrete transform, ANFIS and linear regression for short-term time series prediction of air temperature

This study uses one weather parameter, temperature, contained in the weather station in Bungus port. The used methods are ANFIS, wavelet and statistical models of linear regression. Statistical indicators such as r, RMSE, MAE, and R 2 , are calculated for data analysis. The time series data are used to predict next 60 to 300 minutes for 10 minutes data intervals. Data is shared using Mackey Glass Chaotic Time-Series. Wavelet model is used to analyze sub-time series data. Data input was used the result of wavelet decomposes each level. Approximately 8335 pairs of data are divided into training and testing. Approximate value component level 1 error is smaller than the other levels. ANFIS model with multiple input obtained optimum results during training and testing phase. Linear regression analysis using the model is done by multi-input to give one output result of each prediction. Input data of 60, 120, 180, 240 minutes produces output for data prediction for 300 minutes to determine a combination of the output for multi-input for each minute. The analysis was generated using this model is quite well. However, it is not as well as the ANFIS generated prediction. For the input combination, ANFIS against linear regression is the optimal combination performed on this experiment to this research. The application of computational can be combined with more complex models such as Support Vector Machine, Artificial Neural Network, and hope can step over to next experiment with long term and complex data.
Show more

13 Read more

Prediction of Temperature Time Series Based on Wavelet Transform and Support Vector Machine

Prediction of Temperature Time Series Based on Wavelet Transform and Support Vector Machine

Different scales of some time series are reflected on different frequency bands. Therefore it is not rigorous to predict the trend in the future using SVM regression this processes. The regression on Wavelet coefficients based on SVM on different scales on time series is presented, which gives full consideration of the impact of regularity for series on various scales and frequencies. The shortcoming of high frequency part (short period) weakened by SVM has been overcome. The problem caused by the continuous wavelet transform in the discrete time-series has been solved through the wavelet transform. The monthly temperature time-series data from city Tangshan is applied to the model as an example. The result indicates that the accuracy obtained from SVM method based on wavelet transform is significantly higher than that based on SVM and BP models.
Show more

8 Read more

Electricity Consumption Prediction Using XGBoost Based on Discrete Wavelet Transform

Electricity Consumption Prediction Using XGBoost Based on Discrete Wavelet Transform

In machine learning, many researchers have developed many single or hybrid models for time series prediction, which have been validated in actual data. Ye Ren used a novel hybrid model of integrating the EMD (empirical mode decomposition) and the SVR to predict wind speed in his study, then compared with a variety of hybrid models and finally found that the hybrid model proposed by himself was more accurate [1]. In the study of fault time series prediction, Xin Wang and Ji Wu proposed a hybrid model of singular spectrum analysis and the SVR, compared with various models such as the ARMA and multiple linear regression, and found that the performance of the model in this example was better than those models [2]. Adamowski and Sun compared the relative performance of the coupled wavelet-neural network models (WA–ANN) and regular artificial neural networks (ANN) for flow forecasting at lead times of 1 and 3 days for two different non-perennial rivers in semiarid watersheds of CyprusR, found that the performance of the former was better [3]. Similarly, Venkata Ramana and others haven combined the wavelet transform with ANN to obtain a hybrid model named WNN, and regular ANN and it were then applied to monthly rainfall prediction respectively to gain a conclusion that the prediction accuracy of the latter is better than that of the former [4]. And before that, a scholar has put forward an approach of nonlinear SVM based on PSO (particle swarm optimization algorithm) applied to rainfall forecasting as well [5]. Zhiyong Liu and others investigated a hybrid model that was combined the discrete wavelet transform and support vector regression (the DWT– SVR model) for daily and monthly stream flow forecasting and found it outperformed regular SVR [6]. In addition, some other hybrid models and unicity models for forecasting have also been proposed, such as AGA-SSVR which hybridizes SVR model with adaptive genetic algorithm (AGA) and the seasonal index adjustment [7], Recurrent Neural Networks(RNN) and Grammatical Inference [8], the least squares support vector regression [9,10], SVR or SVM [11,12,13,14], Neural Networks [15]. Most of the above are only developed to short-term time series prediction. Moreover, other investigators have made some results in long-term time series prediction. Alexander Grigorievskiy and others applied OP-ELM to the problem of long-term time series prediction [16]. Multiple-output support vector regression (M-SVR) have been employed in multi-step-ahead time series prediction by Yukun Bao and others [17]. The least squares support vector machine and the multivariate adaptive regression spline model are applied to the long-term prediction of river water pollution by Ozgur Kisi et al [10].
Show more

14 Read more

Effects of Transmit Diversity on a Discrete Wavelet Transform and Wavelet Packet Transform based Multicarrier Systems

Effects of Transmit Diversity on a Discrete Wavelet Transform and Wavelet Packet Transform based Multicarrier Systems

Secondly, MIMO and beam-forming are employed using an antenna array at the transmitter with two elements to provide temporal diversity to the multi-user system. A complete parametric study has been carried out to find the best communication system architecture. The research considers the behavior of different wavelet transforms, families, and filters, and suggest one that has the best reconstruction properties for use in a multicarrier (MC) system. The proposed system consists of three aspects: 1) using wavelets to reduce ICI and multipath fading and to reduce bandwidth usage, 2) using MIMO to ensure high data-rates and 3) beam- forming for good link reliability. Most of the processing burden is kept at the Base-Station (BS) with less cost and power constraints.
Show more

9 Read more

Unified Discrete Wavelet Transform with Ridge Regression and Principal Component Regression to Predict Concentration of Gingerol Compound in Ginger Crop

Unified Discrete Wavelet Transform with Ridge Regression and Principal Component Regression to Predict Concentration of Gingerol Compound in Ginger Crop

Dari 1024 titik yang terpilih dilakukan transformasi wavelet diskret (TWD), dengan melihat berbagai ke- mungkinan resolusi yang menghasilkan koefisien-koe- fisien wavelet yang jumlahnya lebih kecil dari jumlah sampel untuk kelompok data kalibrasi, serta berbagai fungsi mother wavelet keluarga Daubechies. Alasan pemilihan mother wavelet keluarga Daubechies karena sering dipakai dalam aplikasi dan memberikan hasil pemodelan yang baik [8], [9], [10]. Koefisien-koefisien wavelet yang dihasilkan digunakan untuk pengembangan model kalibrasi peubah ganda. Perhitungan matriks wa- velet pada penelitian ini menggunakan software wavetresh 3 seperti yang diterangkan pada referensi [6], [7]. Karena lama penyimpanan berpengaruh terhadap kadar gingerol yang dihasilkan, maka dalam pencarian model prediksi yang lebih baik diikutkan peubah dummy yang mencerminkan kelompok lama penyimpanan (un- tuk data pada penelitian ini 3 bulan dikode 0 dan 10 bu- lan dikode 1).
Show more

6 Read more

Index Terms Discrete Wavelet Transform, Image Compression Haar wavelet transform lossy compression

Index Terms Discrete Wavelet Transform, Image Compression Haar wavelet transform lossy compression

Abstract — Data compression which can be lossy or lossless is required to decrease the storage requirement and better data transfer rate. One of the best image compression techniques is using wavelet transform. It is comparatively new and has many advantages over others. Wavelet transform uses a large variety of wavelets for decomposition of images. The state of the art coding techniques like HAAR, SPIHT (set partitioning in hierarchical trees) and use the wavelet transform as basic and common step for their own further technical advantages. The wavelet transform results therefore have the importance which is dependent on the type of wavelet used .In our thesis we have used different wavelets to perform the transform of a test image and the results have been discussed and analyzed. Haar, Sphit wavelets have been applied to an image and results have been compared in the form of qualitative and quantitative analysis in terms of PSNR values and compression ratios. Elapsed times for compression of image for different wavelets have also been computed to get the fast image compression method. The analysis has been carried out in terms of PSNR (peak signal to noise ratio) obtained and time taken for decomposition and reconstruction.
Show more

8 Read more

A comparison of discrete cosine transform and discrete wavelet transform algorithm in watermarking against common attacks

A comparison of discrete cosine transform and discrete wavelet transform algorithm in watermarking against common attacks

(Nivedita et al., 2012), (Fung et al., 2011). Watermarking based on DCT has two facts. The first fact is that much of the signal energy lies at low-frequencies sub-band which contains the most important visual parts of the image. The second fact is that the high frequency components of the image are usually removed during compression and noise attacks. Therefore, the watermark is embedded by modifying the coefficients of the middle frequency sub-band so that the visibility of the image is not affected and the watermark is not removed by compression (Amirgholipour and Nilchi, 2009). There are many variants of the Discrete Cosine Transform, but DCT that have 2-dimensional is the most commonly used for digital images. The formula of 2D (N by M) DCT is defined as the following:
Show more

37 Read more

Image Compression Using Discrete Cosine Transform (DCT) & Discrete Wavelet Transform (DWT) Techniques

Image Compression Using Discrete Cosine Transform (DCT) & Discrete Wavelet Transform (DWT) Techniques

Abstract: Image compression is a process of reducing or eliminating redundant or irrelevant data. It helps in effective utilization of high speed network resources. Many techniques are available for compressing the images. This paper addressed the two compression techniques, i.e. Discrete Cosine Transform (DCT) and Discrete Wavelets Transform (DWT) that are widely used. DCT is an orthogonal transform and attempts to decorate the image data. DWT is a mathematical tool for changing the coordinate system which represents the signal to another domain that is best suited for compression. The objective is to compare two compression techniques (DCT &DWT) and validate the results using MATLAB. In order to meet the objective the both theoretical and practical approach has been used. The research methodology used practical approach for performance analysis. MATLAB is used as a simulator to implement the techniques of steganography. This is a comparative study based on peak- signal-to-noise (PSNR), compression time (CR) and mean square error (MSE) values of image qualities for corresponding techniques.The performance of the block-based DCT scheme degrades at high compression ratio. On the other hand,the output of the DWT image compression is good.
Show more

10 Read more

ECG Signal  Denoising by Discrete Wavelet Transform

ECG Signal Denoising by Discrete Wavelet Transform

At the end of this work and in accordance with the results obtained in the previous sections, multiple conclusions can be issued. The first conclusion is about the appro- priate wavelet function for ECG denoising. Indeed, the wavelet functions Symlet 8 and Coiflet 4 are to be better more than any other wavelet for the process of removal of EMG and baseline wander. On the other hand, to eliminate PLI, it is recommended to use the Bior 3.5 wavelet function.

18 Read more

Progressive 1-D Discrete Wavelet Transform

Progressive 1-D Discrete Wavelet Transform

The discrete wavelet change (DWT) is as a rule logically utilized for picture coding. It is a direct result of the way that DWT backup highlights like dynamic picture transmission with outrageous quality and determination. The DWT is the basic part of the JPEG2000 framework, [1] and it additionally has been selected as the change coder in MPEG-4 still surface coding. Nonetheless, the DWT calculation or much contrast with that of discrete cosine change (DCT) due to calculations of channel. These days, lifting is broadly picked conspire for DWT usage in light of its points of interest like change in the speed and the general impression measure and less intricate design contrast
Show more

6 Read more

Digital Watermarking using Discrete Wavelet Transform

Digital Watermarking using Discrete Wavelet Transform

Figure 1: 1-level discrete wavelet decomposition Discrete wavelet transform is used in many applications related to signal processing such as compression of audio and video, also used in noise removal. It provides high compression ratio with good quality of reconstruction.

5 Read more

An Efficient FPGA Implementation of the Discrete Wavelet Transform

An Efficient FPGA Implementation of the Discrete Wavelet Transform

A similar technique for coefficient-interleaving was presented by Shahid Masud and John V.McCanny [16] to implement a three level 1-D orthonormal DWT on a Xilinx 4052XL FPGA. The first stage implemented the coefficient-interleaved DF FIR filters. Polyphase techniques were not used for this architecture. However, there was no wastage of computations because the coefficients in the CSRs were alternated depending upon the even and odd parts of the signal. This implementation made use of buffer memory to temporarily store intermediate output values. Each stage differed from the previous stage making the design complex. Also, for higher stages of implementation, this architecture required significant switching activity to correctly synchronize the data and coefficients which could increase the power required for the design. Additionally, the critical path increased with each stage and the throughput was reduced by two after level 2 and by four after level 3 which may make this archi- tecture unsuitable for real time applications.
Show more

102 Read more

A Review of VLSI Architectures for Discrete Wavelet Transform

A Review of VLSI Architectures for Discrete Wavelet Transform

Ferratti and Rizzo [8] presented a dedicated architecture for evaluating 2D DWT using (4,2) filter. The block diagram of the (4,2) filter shown in figure 20, it has two parallel filters which predict and update the values along Rows and two parallel filters which predict and update the values along columns, Buffers are added in order to store the intermediate data generated by parallel filters. Dual ported buffers are used in four data words are accessed at a time. The filter parameters taken for Pred-row is L g =4. P red row is used to compute the „H‟ values. Upd-row uses filter parameter L h = 2 and it also requires „H‟ value to evaluate
Show more

7 Read more

Using Data Mining with Time Series Data in Short Term Stocks Prediction: A Literature Review

Using Data Mining with Time Series Data in Short Term Stocks Prediction: A Literature Review

One important application concerns short-term stocks prediction, which is the main focus of this paper. In [18], an approach to the paradox of obtaining better results with long-horizon forecasts than with short-horizon fore- casts is presented, and it is claimed that the paradox is solved, since the proposed model obtains promising re- sults. Nevertheless, there is a great deal of interest from investors in short-horizon forecasts, thus the authors con- sider that research focusing on this issue is important, namely in using data mining with time series for short- term stocks prediction.
Show more

5 Read more

Boosting wavelet neural networks using evolutionary algorithms for short-term wind speed time series forecasting

Boosting wavelet neural networks using evolutionary algorithms for short-term wind speed time series forecasting

Many practical time series modelling problems can be described as follows. There is a response variable y (also known as output or dependent variable) that depends on a set of independent variables x  { , x x 1 2 ,..., x n } (also known as input or explanatory variables). Usually, a number of observations of both the output and input variables are available, which are denoted by { y x k , k } (k =1, 2,…,N). The true quantitative representation of the relationship between the output y and the input x is in general not known. The central task of data modelling is to establish quantitative representa- tions, e.g. mathematical models such as y = f(x) + e (where e is model error), to ap- proximate the input-output relationship as close as possible.
Show more

13 Read more

Unsupervised Feature Extraction for Time Series Clustering Using Orthogonal Wavelet Transform

Unsupervised Feature Extraction for Time Series Clustering Using Orthogonal Wavelet Transform

dataset and n is the length of each time series in the dataset. DWT and DFT are powerful signal processing techniques, and both of them have fast computational algorithms. DFT maps the time series data from the time domain to the fre- quency domain, and there exists a fast algorithm called Fast Fourier Transform (FFT) that can compute the DFT coeffi- cients in O(mnlogn) time. DFT has been widely used in time series indexing [4, 37, 42]. Unlike DFT, which takes the original time series from the time domain and trans- forms it into the frequency domain, DWT transforms the time series from time domain into time-frequency domain. Since the wavelet transform has the property of time- frequency localization of the time series, it means most of the energy of the time series can be represented by only a few wavelet coefficients. Moreover, if we use a spe- cial type of wavelet called Haar wavelet, we can achieve O(mn) time complexity that is much efficient than DFT. Chan and Fu used the Haar wavelet for time-series classifi- cation, and showed performance improvement over DFT [9]. Popivanov and Miller proposed an algorithm us- ing the Daubechies wavelet for time series classification [36]. Many other time series dimensionality reduction techniques also have been proposed in recent years, such as Piecewise Linear Representation [28], Piecewise Aggre- gate Approximation [25, 45], Regression Tree [18], Sym- bolic Representation [32]. These feature extraction algo- rithms keep the features with lower reconstruction error, the feature dimensionality is decided by the user given ap- proximation error. All the proposed algorithms work well for time series with some dimensions because the high cor- relation among time series data makes it possible to re- move huge amount of redundant information. Moreover, since time series data are normally embedded by noise, one byproduct of dimensionality reduction is noise shrinkage, which can improve the mining quality.
Show more

15 Read more

Discrete Fourier transform based frequency characteristics of iterative learning control for linear discrete time systems

Discrete Fourier transform based frequency characteristics of iterative learning control for linear discrete time systems

Actually, by virtue of the repetition feature of the ILC system, the iteration-wise out- put can be regarded as a segment of a periodic sequence whereas for a periodic sequence, the DFT is a well-known powerful technique, which expresses the sequence with the fun- damental N period as a summation of a fundamental sine wave plus (N – 1) harmonic sine waves. Its equivalent form is in exponential complex-variable functions. Then the frequency-domain spectrum can be precisely computed by DFT in a direct manner. This motivates the paper firstly to investigate DFT properties for linear discrete time-invariant (LDTI) derivative-type (D-type) ILC systems and the convergence of the Toeplitz matrix to the power of the iteration index. The followed works are the frequency-domain con- vergence derivation of LDTI D-type ILC systems and its generalization to linear discrete time-varying (LDTV) D-type ILC systems.
Show more

22 Read more

Medical Image Fusion Using Discrete Wavelet Transform

Medical Image Fusion Using Discrete Wavelet Transform

Medical image fusion is the process of registering and combining multiple images from single or multiple imaging modalities to improve the imaging quality and reduce randomness and redundancy in order to increase the clinical applicability of medical images for diagnosis and assessment of medical problems. Multimodal medical image fusion algorithms and devices have shown notable achievements in improving clinical accuracy of decisions based on medical images. The domain where image fusion is readily used nowadays is in medical diagnostics to fuse medical images such as CT (Computed Tomography), MRI (Magnetic Resonance Imaging) and MRA. This paper aims to present a new algorithm to improve the quality of multimodality medical image fusion using Discrete Wavelet Transform (DWT) approach. Discrete Wavelet transform has been implemented using different fusion techniques including pixel averaging, maximum minimum and minimum maximum methods for medical image fusion. Performance of fusion is calculated on the basis of PSNR, MSE and the total processing time and the results demonstrate the effectiveness of fusion scheme based on wavelet transform. KEYWORDS : Image Fusion, Multimodal medical image fusion, fusion rules, PSNR, MSE.
Show more

6 Read more

Bearing fault detection using discrete wavelet transform

Bearing fault detection using discrete wavelet transform

For normal speeds, these defect frequencies are usually less than 500 Hz (Tandon and Choudhury, 1999). However, these frequencies may be slightly different from values calculated as there are other external factors that influence the results such as slipping or skidding in the rolling element bearings (Prasad, 1987). An example of typical spectrum due to an inner race defect is shown in Figure 2.2. The sidebands have been attributed to the time-related changes in defect position relative to the vibration measuring position (Igarashi and Hamada, 1982). Tandon and Choudhury in 1999 manage to derive an expression for frequencies and relative amplitudes of the various spectral lines based on the flexural vibration of races due to a localized defect on one of the bearing elements.
Show more

24 Read more

Identification Of Faults And Its Location For The Series Compensated Transmission Line Using Discrete Wavelet Transform

Identification Of Faults And Its Location For The Series Compensated Transmission Line Using Discrete Wavelet Transform

This is a progressive work of my previous paper [1]. Distance relay protecting series compensated line has limitation with metal oxide varistor (MOV) operation, high resistance fault, prefault system condition and shunt capacitance. In this paper, a novel technique for the protection of transmission lines for line with fixed series compensation connected at one end using local measurement is proposed. The proposed system uses Discrete Wavelet Transform (DWT) which is widely used in recent times for power system protection. DWT is used here to extract the unseen factors from the fault signals by performing decomposition at different levels. Daubechies wavelet “dB5” is used with single level decomposition and adaptive threshold is calculated to discriminate and detect the faulty phase. The spot of faults is carried out by obtaining the local fault information and remote location fault information along with the transmission line length. The system is independent of any statistical system data and has negligible fault resistance. Test system is modelled in EMTP and fault signals are generated to test the reliability of the algorithm. The projected system promises the result by detecting, classifying and locating all the ten faults possible in the transmission line of the power system. A variety of fault cases and simulation results are introduced to show the effectiveness of such algorithm
Show more

7 Read more

Show all 10000 documents...