• Predicting Network Traffic using Radial-basis Function Neural Networks

    Predicting Network Traffic using Radial-basis Function Neural Networks – Fractal Behavior

    I found a paper about Predicting Network Traffic using RBFNN.  I wrote this back in December 2011 regarding Radial-basis Function Neural Networks (RBFNN). Currently, new trends in artificial intelligence are key and RBF-Kernels are in use by machine learning methods and systems.

    ” Fractal time series can be predicted using radial basis function neural networks (RBFNN). We showed that RBFNN effectively predict the behavior of self-similar patterns for the cases where their degree of self-similarity (H) is close to the unity. In addition, we observed the failure of this method when predicting fractal series when H is 0.5. ”


    As Hurst-parameter is closer to 0.5 then RBFF are useless to predict fractal behavior, as shown, the randomness of a Hurst parameter at 0.5

    For the BRW (brown noise, 1/f²) one gets

             Hq = ½,

    and for pink noise (1/f)

            Hq = 0.


    Obviously, the Hurst Parameter or Hurst Exponent is nothing but a degree of  “fractality” for a data set. In General, we don’t expect to predict noise, there is no practical use of for this particular case.  we are using the Hurst parameter to see when the RBFF is capable of finding a right response to the data being captured or introduced to the set.


    Fractal time series can be predicted using RBFNN when the degree of self-similarity, Hurst parameter, is around 0.9. The mean square error (MSE) of the real and predicted sequences was measured to be 0.36 as a minimum. Meanwhile, fractal series with H=0.5 cannot be predicted as well as the ones with higher values of H.

    It was expected that due the clustering process, a better approximation could be achieved using a greater value of M and small dimensionality, however behavior was not observed and in contrast, the performance had and optimal point at M=50 using d=2. This phenomena would require a deeper study and it is out of the scope of this class report.

    Future Work

    I am retaking this work and combining it with all BigData, this should be co-rrelated with RF and other systems and related research.

    Introduction Big Data in RF Analysis | Hadoop: Tutorial and BigData 


    Fractal time series can be predicted using radial basis function neural networks (RBFNN). We showed that RBFNN effectively predict the behavior of self-similar patterns for the cases where their degree of self-similarity (H) is close to the unity. In addition, we observed the failure of this method when predicting fractal series when H is 0.5.


    We will first review the meaning of the term fractal. The concept of a fractal is most often associated with geometrical objects satisfying two criteria: self -similarity and fractional dimensionality. Self- similarity means that an object is composed of sub-units and sub-sub-units on multiple levels that (statistically) resemble the structure of the whole object. Mathematically, this property should hold on all scales. However, in the real world, there are necessarily lower and upper bounds over which such self-similar behavior applies. The second criterion for a fractal object is that it has a fractional dimension. This requirement distinguishes fractals from Euclidean objects, which have integer dimensions. As a simple example, a solid cube is self-similar since it can be divided into sub-units of 8 smaller solid cubes that resemble the large cube, and so on. However, the cube (despite its self- similarity) is not a fractal because it has an (=3) dimension. [1]

    The concept of a fractal structure, which lacks a characteristic length scale, can be extended to the analysis of complex temporal processes. However, a challenge in detecting and quantifying self-similar scaling in complex time series is the following: Although time series are usually plotted on a 2- dimensional surface, a time series actually involves two different physical variables. For example, in Figure 1. the horizontal axis represents “time,” while the vertical axis represents the value of the variable that changes over time. These two axes have independent physical units, minutes and bytes/sec respectively (For example). To determine if a 2-dimensional curve is self-similar, we can do the following test: (i) take a subset of the object and rescale it to the same size of the original object, using the same magnification factor for both its width and height; and then (ii) compare the statistical properties of the rescaled object with the original object. In contrast, to properly compare a subset of a time series with the original data set, we need two magnification factors (along the horizontal and vertical axes), since these two axes represent different physical variables.

    Fig. 1. Fractal time series

    In the different windows observed, hand h2, we can observe a linear dependency between the variances and windows sizes. In other words, the slope is determined by (log(s2) – log(s1))/(log(1) – log(h2)). This slope value is also called Hurst parameter (H) and in general a value of 0.5 indicates a completely brownian process, whereas 0.99 indicates highly fractal.

    The research conducted by Sally Floyd and Vern Paxon [2] concluded that network traffic is fractal in nature and H>0.6. Therefore, RBFNN could be used in this field for network traffic control and analysis. Indeed, we made use of Vern Paxson’s [3,4] method to generate a fractal trace based upon the fractional gaussian noise approximation. The inputs of the Paxson’s program developed are: media, variance, Hurst parameter, and the amound of data. We decided to maintain a media at zero, = 0, and the variance, σ2=1, and 65536 points. Figures 2 and 3, depict the fractal time series at different sampling windows

    Fig. 2. Fractal sequence sampled at different intervals H=0.5, =0 and s2 = 1

    Fig 2. depicts the generated sampled used for training and testing of the GBRF. The signal is composed of 65536 data samples, ranging between 4 and –4, although we only used 10000 points for training and 10000 points for testing. Similarly, Fig. 3 presents the histogram and fast Fourier transform corresponding to the input in Fig. 2.

    Fig. 3. Histogram and fast Fourier transform of the self-similar sequence H=0.5

    Fig. 4. Fractal sequence sampled at different intervals H=0.9, =0 and s2= 1

    In addition, Fig 4 and Fig 5 show the input data at H=0.9. Both plots show a big difference in the frequency domain among the time series with different values of H. This difference allow us to speculate that RBFNN will be able to perform much better than in the purely random case.

    Fig. 5 Histogram and fast Fourier transform of the self-similar sequence H=0.9

    Radial basis functions

    A radial basis function, like an spherical Gaussian, is a function which is symmetrical about a given mean or center point in a multi-dimensional space [5]. In the Radial Basis Function Neural Network (RBFNN) a number of hidden nodes with radial basis function activation functions are connected in a

    feed forward parallel architecture Fig 6.. The parameters associated with the radial basis functions are optimized during training. These parameter values are not necessarily the same throughout the network nor directly related to or constrained by the actual training vectors. When the training vectors are presumed to be accurate ie. Non-stochastic, and it is desirable to perform a smooth interpolation between them, then a linear combination of radial basis functions can be found which gives no error at the training vectors. The method of fitting radial basis functions to data, for function approximation, is closely related to distance-weighted regression. As the RBFNN is a general regression technique it is suitable for both function mapping and pattern recognition problems.

    Fig. 6. Radial basis function representation with k-outputs, M-clusters and d-inputs.

    The equation required by a Gaussian radial basis function (GRBF) equations are shown as follows:

    In all cases, ∈ {1, .. ,N}, or the number of patterns, while ∈ {1, .., K} or the number of outputs, and ∈ { 1, .., M} or the number of clusters used on the network.

    According to Bishop [6] the solution for the weight matrix is defined as follows:

    = Φ(Φ)+ T

    where all these matrices are defined by:

    = {Wkj }

    Φ= {Φ nj }, and Φ nj Φ (xn)

    = {Tnk }, Tnk nk

    And, finally, Y={Ynk }, Y=ΦWT

    Therefore, the weight matrix can be calculated with the formula:

    W Φ +T

    Since Φ is a non-squared matrix the pseudo inverse is required to calculate the matrix W.

    RBFNN and radial basis functions implemented.

    The input to the MATLAB code match up to a file generated by the fractal generator. The set of input data of the fractal file had to be rearranged and organized such that the number of inputs, D, stimulated M- GRBFs. The element D+1 of the sequence was considered as the output. Hence, each sequence of D-inputs will produce one output, which is can be arranged as follows:

    {xi }= {{x[n], x[1], x[ 2], x[ 3],…, x[ D]}

    This set {x i} determines the output tk, which is x[n+1]. This output is used for training of the RBFNN.

    Each term on the {xiinputs, generates a set of mI and j input values, where { 1..M}, and { 1.. N/ (D-1)}. The data is subdivided in N/(MX(D-1)) clusters of D-dimension from which mj and j are calculated. This calculation was done at the cluster of data by first sorting the data according to tkn or the expected outcome. By sorting the (xn} via tkn we will be able to cluster the input hence each independent basis function will represent a cluster of inputs which can generate a similar outcome.

    Hence, it would be expected to have a better predictable value for bigger values of M, or by decreasing the granularity of the cluster. For instance, with d=2, and M=100, given an training set N=3000, we will have a cluster j=1

    The cluster of size 10 will have a mand m2,, which are the medias of the 10 elements in the first and second columns respectively. The variance is determined using all the elements in the cluster, or both columns are rows. Hence, it would be expected that for a big cluster, or a small value of and a high- dimensionality this method lead to bigger error during the approximation.

    Results and experimental prediction using radial basis functions.

    Once the {Φ} matrix is determine, as well as the weight vector , we proceeded to test the RBFNN with some input data.


    Table 1. Variation of and the mean square error of the training sample at different Hurst parameters

    Degree of Self- similarity

    Hurst Parameter


    M- GBRFs



































































    We made use of a sequence, as big as the training input (10000 points). Table 1 depicts the results of the mean square error at different degrees of self-similarity as well as the number of hidden nodes or basis functions used (M).

    Fig 7. Error and comparison between predicted and real sampled signal for d=2 and M=50. Input signal for H=0.9, 10000 samples used for training

    All the input sequences were compared between the original and the predicted input. The best prediction and smaller mean square error (MSE) was observed with d=2, M=50 and H=0.9. This

    behavior can be shown also in the qualitative shape shown in Fig. 7. Where the predicted and real sampled data are very similar and the predicted data follows the real sequence. Although the magnitudes are missing, the RBFNN was able to produce a nice input.

    Fig.8 . Error and comparison between predicted and real sampled signal for d=16 and M=200. Input signal for H=0.9, 10000 samples used for training

    Meanwhile, Fig. 8, shows the results obtained with H=0.9, M=200, d=16 where we observe that there is

    over-estimation on the predicted sequence, which makes the error grow significantly. Those over estimations are not plotted in the figure but rounded between 10 to 20 in magnitude.

    Notwithstanding, the MSE seems to grow to unreasonable values, qualitatively the shape of the predicted sequence follows the real testing sample data.

    Fig 9. Error and comparison between predicted and real sampled signal for d=2 and M=20. Input signal for H=0.5, 10000 samples used for training

    Besides the test executed to the input sequence with H=0.9, Fig. 9 and Fig. 10, depict the behavior of the RBFNN under H=0.5 stimulation. Both plots show the poor performance of the RBFNN when this type of stimulation was employed. In fact, Table 1, presents that the minimum MSE was of 0.5, whereas with H=0.9 the minimum was around 0.3. We have to clarify that for each data set used the RBFNN was trained and its weight matrix calculated using a set of the same input pattern. The performance of the neural network was tested using a training pattern using the same as in the training set.

    As show in Fig. 10, the worst performance of the RBFNN was observed when using 16 inputs (d=16) to determine the a predicted pattern and M=100. Although, the error is higher than the MSE measured in Fig 9, qualitatively this shape seems to follow the real sequence used as input.

    Fig 10. Error and comparison between predicted and real sampled signal for d=16 and M=100. Input signal for H=0.5, 10000 samples used for training


    Fractal time series can be predicted using RBFNN when the degree of self-similarity, Hurst parameter, is around 0.9. The mean square error (MSE) of the real and predicted sequences was measured to be 0.36 as a minimum. Meanwhile, fractal series with H=0.5 cannot be predicted as well as the ones with higher values of H.

    It was expected that due the clustering process, a better approximation could be achieved using a greater value of M and small dimensionality, however behavior was not observed and in contrast, the performance had and optimal point at M=50 using d=2. This phenomena would require a deeper study and it is out of the scope of this class report.



    % Fractal sequence processing
    % © 2001 – Edwin Hernandez
    % selfSimilar = input (‘ Input the name of the file with self-similar
    % content ‘);
    load selfSimilarH05; x_1 = 1:10;
    y_1 = size(10); x_2 = size(100); y_2 = size(100); x_3 = size(1000); y_3 = size(1000); x_4 = size(10000); y_4 = size(10000); j_1=1;
    j_2=1; j_3=1; j_4=1;
    for i=1:10000,
    if (mod(i, 1000) == 0) x_1(j_1) = i;
    y_1(j_1) = selfSimilarH05(i); j_1 = j_1 + 1;
    if (mod(i, 100) == 0) x_2(j_2) = i;
    y_2(j_2) = selfSimilarH05(i); j_2 = j_2 + 1;
    if (mod(i, 10) == 0) x_3(j_3) = i;
    y_3(j_3) = selfSimilarH05(i); j_3 = j_3 + 1;
    x_4(i) = i;
    y_4(i) = selfSimilarH05(i); end

    subplot(2,2,1); plot(x_1, y_1);
    title(‘Sampled at 1000 sec’, ‘FontSize’, 8 );
    %xlabel(‘time (s)’,’FontSize’, 8 );
    ylabel(‘Data’,’FontSize’, 8 );

    subplot(2,2,2); plot(x_2, y_2);
    title(‘ Sampled at 100 sec’,’FontSize’, 8 );
    %xlabel(‘time (s)’,’FontSize’, 8 );
    ylabel(‘Data ‘,’FontSize’, 8 );

    subplot(2,2,3); plot(x_3, y_3);
    title(‘ Sampled at 10 sec’,’FontSize’, 8 ); xlabel(‘time (s)’,’FontSize’, 8 );
    ylabel(‘Data’,’FontSize’, 8 );
    subplot(2,2,4); plot(x_4, y_4);
    title(‘ Sampled at 1 sec’,’FontSize’, 8 ); xlabel(‘time (s)’,’FontSize’, 8 );
    ylabel(‘Data’,’FontSize’, 8 );

    subplot(2,1,2), plot(log(abs(fft(y_4, 1024))));
    title(‘ Fast fourier transform (1024 samples)’,’FontSize’, 8); ylabel(‘log10’, ‘FontSize’, 8);
    xlabel(‘frequency domain’, ‘FontSize’, 8);

    subplot(2,1,1), hist(y_4,100); xlabel(‘Data in 100 bins’,’FontSize’,8); ylabel(‘Samples’,’FontSize’, 8);
    title(‘ Histogram ‘,’FontSize’, 8);

    pause H=20
    for i=1:H-1,
    x(i) = size(round(10000/H)); end
    yk = size(round(10000/5));
    % 4 y 1 output to create Yk samples j=1;
    load selfSimilarH09; for i=1:H:10000,
    for k=0:H-2,
    x1(j) = selfSimilarH09(i+k); end
    yk(j) = selfSimilarH09(i+k+1); j=j+H;

    subplot(5,1,1), plot(x1);
    subplot(5,1,2), plot(x2);
    subplot(5,1,3), plot(x3);
    subplot(5,1,4), plot(x4);
    subplot(5,1,5), plot(yk);
    % Gaussian radial basis functions
    % ——————————————————————–
    % Edwin Hernandez
    % Modified to sort the clusters and then find the Mu’s and the sigmas.
    % if M=100 I will sort all the clusters in 100 piles.

    load selfSimilarH09 NDATA = 10000;
    % get all the chunks and the T matrix
    % out of all the inputs only 65500 I’ll use k=1;
    x = size(round(NDATA/(D+1)), D); t = size(round(NDATA/(D+1)));

    for j=1:round(NDATA/(D+1)), for i=1:D,
    x(j,i) = selfSimilarH09(k);
    k=k+1; end k=k+1;
    t(j) = selfSimilarH09(k); end

    u = size(size(x), D+1); u = [x, t’];
    u = sortrows(u, D+1);

    x = u(1:size(x), 1:D);
    t = u(C*R+1:(C+1)*R)’;
    %cwd = pwd;

    L = size(t);
    cluster = floor(R*C/(M*D)); Mu = size(M, D);
    sigma = size(M); Mean = size(D,1); k=0;
    for j=1:M,
    if (j<M)
    z= x(k+1:j*cluster,1:D); k=j*cluster; [l,c]=size(z);
    sigma(j) = cov(z(1:l*c)); Mean = mean(z);
    %pause; else
    z= x(k+1:R,1:D);
    k=j*cluster; [l,c]=size(z);
    sigma(j) = cov(z(1:l*c));
    Mean =mean(z);
    for i=1:D, Mu(j,i)=Mean(i);
    end end
    cwd = pwd;
    cd(tempdir); pack

    Phi = size(M, round(NDATA/(D+1))); % M, GBRF ….
    for j=1:M,
    for k=1:round(NDATA/(D+1)), dist = 0;
    for i=1:D,
    dist = dist + (x(k, i) – Mu(j, i))^2; end
    Phi(j, k) = exp( -2*dist/(2*sigma(j))); end

    cwd = pwd; cd(tempdir); pack

    % Weight matrix . W = size(M, 1);
    W = pinv(Phi)’*t;

    x_test = size(round(NDATA/(D+1)), D); t_test = size(round(NDATA/(D+1)));

    for j=1:round(NDATA/(D+1)), for i=1:D,
    x_test(j,i) = selfSimilarH09(k); k=k+1;
    end k=k+1;
    t_test(j) = selfSimilarH09(k); end

    error = size(round(NDATA/(D+1))); y = size(round(NDATA/(D+1))); Phi_out = size(M);
    meanSQRerror = 0
    for k=1:round(NDATA/(D+1)), for j=1:M,
    dist = 0; for i=1:D,
    dist = dist + (x_test(k, i) – Mu(j, i))^2; end
    Phi_out(j) = exp( -2*dist/(2*sigma(j)));
    y(k) = Phi_out*W;
    error(k) = y(k) – t_test(k);
    meanSQRerror = 0.5*(y(k)-t_test(k))^2+meanSQRerror;
    if abs(y(k))>=5 y(k) = 5;

    if abs(error(k))>=5 error(k)=5;


    fprintf(‘The mean square error is : %f’, meanSQRerror); c=round(NDATA/(D+1));
    subplot(2,1,1), plot(1:c, error);
    title(‘ Prediction error ‘,’FontSize’, 8);
    %subplot(3,1,2), hist(error, 100);
    %title(‘ Error histogram ‘, ‘FontSize’, 8); subplot(2,1,2), plot(1:c, t_test(1:c), ‘r:’,1:c, y); title(‘ Real and predicted Data ‘, ‘FontSize’, 8); legend(‘Real’,’predicted’);


    1. Peng C-K, Hausdorff JM, Goldberger Fractal Analysis Methods http://reylab.bidmc.harvard.edu/tutorial/DFA/node1.html
    2. Vern Paxson and Sally Floyd, Wide-Area Traffic: The Failure of Poisson ModelingIEEE/ACM Transactions on Networking, Vol. 3 No. 3, pp. 226-244, June 1995.
    3. Vern Paxson. Fast Approximation of Self-Similar Network Traffic. Technical Report LBL-36750, Lawrence Berkeley Labs, April 1995.
    4. Vern Paxson. http://ita.ee.lbl.gov/html/contrib/fft_fgn_c-readme.txt
    5. Radial Basis Functions. http://www.maths.uwa.edu.au/~rkealley/ann_all/node162.html
    6. Christopher Bishop “Neural networks for pattern recognition”, Oxford University Press, Birmingham, UK, 1995.


    [spiderpowa-pdf src=”http://edwinhernandez.com/wp-content/uploads/2017/08/GRBFF.pdf”]GRBFF



  • Introduction Big Data in RF Analysis

    Big Data in RF Analysis

    Big Data provides tools and a framework to analyze data, in fact, large amounts of data. Radio Frequency, RF, provides  amounts of information that depending n how it is modeled or created, its analysis fits many statistical models and is in general  predicted using passive filtering techniques.

    The main tools for Big Data include statistical aggregation functions,  learning algorithms, and the use of tools. There are many that can be purchased but many that are free but may require certain level of software engineering.  I love Python and specially the main modules used in python are:

    • Pandas
    • SciPy
    • NumPy
    • SKLearn

    and, there are many more used for the analysis and post-processing of RF captures.

    Drive Test and Data Simulation

    In general, many drive test tools are used to capture RF data form LTE/4G, and many other systems. As vendors, we can find Spirent, and many others, and we can capture RF from multiple base stations and map those to Lat/Long in a particular area covered by many base stations.  It’s obvious that drive test cannot cover the entire area, as  expected extrapolation and statistical models are required to complete the drive test.

    In a simulator, just as in MobileCDS and other simulators, specially those in “Ray Tracing,” the simulator uses electromagnetic models to compute the RF received by an antenna.


    Big Data Processing for a Massive Simulation

    Unstructured data models are loaded with KML and other 3D simulation systems that include polygons and buildings that are situated on top of a google earth map or any other map vendor.  The intersection of the model with the 3D database produces the propagation model that needs massive data processing, Map-Reduce and Hadoop to handle the simulation.

    HADOOP and MAP Reduce for RF Processing

    The data is then stored in unstructured models with RF information, that include the Electromagnetic field, frequency, time, delay, error, and other parameters that are mapped to each Lat/Log or x,y, z coordinates in the plane being modeled.  The tools are usually written in Python and parallelization can be done in multiple hadoop nodes and processing of CSV/TXT files with all the electromagnetic data and the 3D map being rendered.


    As you can see the Hadoop/GlusterFS is our choice, as we don’t see that much value for HDFS or the Hadoop Data File System are the ones that handle all the files and worker systems.  As you can tell, we are fans of GlusterFS and processing of all Hadoop cluster nodes is managed in a massive processing network of high-performance networks and 10Gb Fiber network.

    Big Data models: OLTP and OLAP  Processing

    The OLTP and OLAP data models definitions can be found online:

    ” – OLTP (On-line Transaction Processing) is characterized by a large number of short on-line transactions (INSERT, UPDATE, DELETE). The main emphasis for OLTP systems is put on very fast query processing, maintaining data integrity in multi-access environments and an effectiveness measured by number of transactions per second. In OLTP database there is detailed and current data, and schema used to store transactional databases is the entity model (usually 3NF).


    – OLAP (On-line Analytical Processing) is characterized by relatively low volume of transactions. Queries are often very complex and involve aggregations. For OLAP systems a response time is an effectiveness measure. OLAP applications are widely used by Data Mining techniques. In OLAP database there is aggregated, historical data, stored in multi-dimensional schemas (usually star schema). “


    We have different research areas:

    • Analysis of data for handover protocols,
    • Data mining for better antenna positioning,
    • Machine learning techniques for better PCRF polices and more




  • Big Data Presentation and 4G/5G Intelligent Cell Positioning

    BigData Presentation – Radio Frequency / Mobile CDS
    Intelligent Positioning of RF Cells for 4G/5G

    I was invited to FAU (Florida Atlantic University) to present at one of the MBA classes on “Big Data Analytics” and we went over the important concept and examples of MapReduce, Hadoop, Pandas, and sample on how Radio Frequency can be simulated and how Big Data is the key component to process, aggregate, and create dashboards of RF simulations over 3D KML maps loaded from Google Earth/Google Maps. This presentation also covered aspects on how the data is split and can be splitter in multiple GPUs using OpenCL as a framework.




    You can see that yourself here: http://4Gexpert.com/