Sun, J. Zhang, P. Rimba, S. Gao, Y. Xiang, L. Zhang, Data-driven cybersecurity incident prediction: A survey.

### References

IEEE Commun. Bou-Harb, P. Denning, An intrusion-detection model. SE 2 , — Markou, S. Singh, Novelty detection: a review part 1: statistical approaches. Chandola, A. Banerjee, V. Kumar, Anomaly detection: a survey. ACM Comput. Neil, C.

Hash, A. Brugh, M. Fisk, C. Storlie, Scan statistics for the online detection of locally anomalous subgraphs. Deng, D. Yu, et al.

## A Tour of Machine Learning Algorithms

Karlsson, A. Loutfi, A review of unsupervised feature learning and deep learning for time-series modeling. Pattern Recogn. Cavalcante, R. Brasileiro, V. Souza, J. Nobrega, A. Oliveira, Computational intelligence and financial markets: A survey and future directions. Expert Syst. Li, Q. Li, Y. Ye, S. Li, R. Baral, T. Li, H. Wang, Q. Li, S. Xu, Hashtran-dnn: a framework for enhancing robustness of deep neural networks against adversarial malware samples.

Li, D. Zou, S. Xu, X. Ou, H. Jin, S. Wang, Z. Deng, Y. Vuldeepecker: A deep learning-based system for vulnerability detection Internet SocietySan Diego, Xu, H. Jin, Y. Zhu, Z. Chen, S. Wang, J. Wang, Sysevr: A framework for using deep learning to detect software vulnerabilities. Grieco, G. Grinblat, L. Uzal, S.

Rawat, J. Feist, L. Jin, H. Qi, J. Vulpecker: an automated vulnerability detection system based on code similarity analysis ACMLos Angeles, , pp.

- Financial Time Series Prediction Using Deep Learning;
- Co pilot Live Android - Copilot For Android?
- ESANN - proceedings | ESANN ?
- Oracle BAM 11gR1 Handbook.
- 1 INTRODUCTION.
- Multivariate Time Series Forecasting with LSTMs in Keras.

Chen, M. Khandaker, Z. Cryer, K. Brockwell, R. Ke, H. Zheng, H. Yang, X. Chen, Short-term forecasting of passenger demand under on-demand ride services: A spatio-temporal deep learning approach. C Emerg. Barabas, G. Boanea, A. Rus, V. Dobrota, J. Evaluation of network traffic prediction based on neural networks with multi-task learning and multiresolution decomposition IEEECluj-Napoca, , pp.

Azzouni, G. Siami-Namini, A. Kuan, T. Liu, Forecasting exchange rates using feedforward and recurrent neural networks. Mikolov, M. Burget, J. Sundermeyer, I. Oparin, J. Gauvain, B. Freiberg, R. Huang, G.

## Subscribe to RSS

Zweig, B. Cache based recurrent neural network language model inference for first pass speech recognition IEEEFlorence, , pp. Liu, Y. Wang, X. Gales, P. Schuster, K. Paliwal, Bidirectional recurrent neural networks. Bengio, P. Simard, P. Frasconi, Learning long-term dependencies with gradient descent is difficult. Neural Netw. Hochreiter, J. Schmidhuber, Long short-term memory. Neural Comput. Goodfellow, Y. Bengio, A.

Kingma, J. Ba, Adam: A method for stochastic optimization. Hyndman, A. Koehler, Another look at measures of forecast accuracy. Baecher, M. Koetter, T. Holz, M. Dornseif, F. The nepenthes platform: An efficient approach to collect malware SpringerBerlin, Heidelberg, , pp. Almotairi, A. Clark, G. Mohay, J. Zhang, Time series forecasting using a hybrid arima and neural network model.

Kumar, M. Thenmozhi, Forecasting stock index returns using arima-svm, arima-ann, and arima-random forest hybrid models. Friedman, T. Hastie, R. Tibshirani, The Elements of Statistical Learning, vol. Pai, C. Lin, A hybrid arima and support vector machines model in stock price forecasting.

Chen, B. Yang, J. Dong, A. Abraham, Time-series forecasting using flexible neural tree model. Download references. Data used in this work is not suitable for public use. XF constructed the deep learning framework and performed the deep learning experiments.

MX and PZ performed the experiments on the statistical models. SX drafted the manuscript. All authors reviewed the draft. All authors read and approved the final manuscript. Correspondence to Xing Fang. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

### Future Predictions

Reprints and Permissions. Search all SpringerOpen articles Search. Abstract Like how useful weather forecasting is, the capability of forecasting or predicting cyber threats can never be overestimated. Our contributions The contribution of the present paper is in two-fold. Related work Statistical methods have been widely used in the context of data-driven cyber security research, such as intrusion detection [ 15 — 18 ]. A standard unfolded RNN structure at time t.

Full size image.

**click**

## Generative model

Data collection The dataset we analyze is the same as the dataset analyzed in [ 1 ]. Data preprocessing As in [ 1 ] and many analyses, we treat flows rather than packets as attacks, while noting that flows can be based on the TCP or UDP protocol. Model training and selection In the training process, we use the mini-batch gradient descent method to compute the minimum of the objective function, which is described in Eq.

Table 4 Prediction accuracy of the selected model with respect to each dataset Full size table. Notes 1. References 1 Z. Article Google Scholar 2 E. Google Scholar 3 S. Google Scholar 4 S. Google Scholar 5 L. Google Scholar 6 C. Google Scholar 7 D. Google Scholar 8 Z. Google Scholar 9 Z. Article Google Scholar 10 C. Article Google Scholar 15 D. Article Google Scholar 16 M. Article Google Scholar 17 V. Article Google Scholar 18 J. Article Google Scholar 21 R.

Article Google Scholar 22 D. Google Scholar 25 Z.

Google Scholar 27 Z. Google Scholar 28 Y. Such models are based on minimizing some prediction error over the training set and are therefore, conceptually, much simpler than a hidden state based model. What would speak in favour of the HMM and what in favour of a regression approach? Intuitively one should take the simpler model possible to avoid over-fitting; this speaks in favour of a stateless approach We also have to consider that both approaches get the same input data for training I think this implies that if we do not incorporate additional domain knowledge in the modelling of a hidden state model, e.

At the end one can of course play with both approaches and see what performs better on a validation set, but some heuristics based on practical experience might also be helpful Perhaps this has an implication for the modelling choice. State-space model hidden state model and other stateless model you mentioned are going to discover the underlying relationship of your time series in different learning paradigm: 1 maximum-likelihood estimation, 2 Bayes' inference, 3 empirical risk minimization.

When you use Baum-Welch to estimate the parameters, you are in fact looking for a maximum-likelihood estimate of the HMM. If you use Kalman filter, you are solving a special case of Bayesian filter problem which is in fact an application of Bayes' theorem on update step :.

On the other hand, for other stateless model you mentioned, like SVM, splines, regression trees, nearest neighbors. Sign up to join this community. The best answers are voted up and rise to the top. Home Questions Tags Users Unanswered. Hidden state models vs. Asked 5 years, 5 months ago. Active 5 years, 4 months ago.

Viewed 1k times. Mannaggia Mannaggia 91 5 5 bronze badges. Dynamic linear regression models in which previous values of the predictand are included on the right-hand side of the model equation would seem very much to be state-conditioned. But perhaps I am missing something. I would say it is a bit a question of semantics, I also give an example of regression models which include the n-past observation values on the right hand side of the model, such a model is of course dynamic.