Abstract— Investors take huge risk in making investments becausethey lack the information about the future sale of the company or we can alsosay that they lack forecasting. Due to which they may have to face loss infuture. Similar is the case with transportation, especially when it comes toairline mode because of huge increase in investment and risk factor. Keeping inmind the above discussed issue this paper deals with forecasting. The paperproposes the use of Long Short-Term Memory Network. Unlike the existingschemes, the proposed scheme is capable enough to minimize the Root Mean SquareError to maximum extent which ultimately would lead to better forecastingresults.

The proposed scheme consist of four steps LSTM model construction,data processing and making data suitable for LSTM model, fitting a statefulLSTM network model to the training data and at last we evaluate the static LSTMmodel on the test data and report the performance of forecast .Furthermore, theobtained results indicates that the proposed scheme reduces the chances offalse positive to great extent and is practical enough to be implemented inreal-time scenario. Keywords- LSTM , Time series prediction ,Forecasting                                                                                                                                                   I.      IntroductionTime series modeling is a dynamic research area which hasattracted attentions of researcher’s community over last few decades. The mainaim of time series modeling is to carefully collect and rigorously study thepast observations of a time series to develop an appropriate model whichdescribes the inherent structure of the series. This model is then used togenerate future values for the series, that is to make forecasts. Time seriesforecasting thus can be termed as the act of predicting the future byunderstanding the past 1.

Now a days travelling is playing most important role inlife of people because of which huge amount of investment is made by variouspeople. Faster you travel from one place to another better it is for user andbecause of which airlines are major targets to investors. This paper deals withthe airline dataset which gives the current and previous records of number ofpassengers travelled using airline.

With help of the theses records this paperwill analyse the number of passengers in the upcoming months and years. Thisinformation will be the key factor for investors to decide whether they want toinvestment in the airline business or not depending the change in rate of thenumber of customers whether its positive or negative.For solving the above problem we have used StackedStateful LSTM with memory between the batches. In the proposed schema we willconstruct a LSTM model. LSTM is the type of Recurrent Neural Network (RNN).

Thebenefit of this type of network is that it can learn and remember over longsequences and does not rely on a pre-specified window lagged observation. Thedataset loaded contains information of the number of passengers who had travelledthrough the airline in particular month and year. The dataset needs to betransformed to make it more suitable for LSTM model which include.  A. MOTIVATION Risk factor for any investment increases as the amount ofinvestment increases because of which investors do keep tract of the company’sprofile and measure the stability of the company. But predictions from theabove method turn out of to give huge amount of error. This paper proposed ascheme to reduce this error and for more accurate forecasting.

The majormotivation of this paper is to reduce the risk factor and also to analyse thecompany. B. CONTRIBUTION The primary contribution of this paper is to LongShort-Term Model is to that it reduces the Root Mean Square Error of LSTM modelwith use of Keras deep learning library. The another major contribution of thispaper is the combination of Keras deep learning library with LSTM.  C. ORGANIZATION The rest of the paper is organized as follows. Section IIgives the relative work which are done in this field.

  Section III gives the brief description aboutthe working of the proposed scheme. Section IV elaborates the LSTM model. Theresults and discussion are presented in Section V. The paper is finallyconcluded in Section V.                                                                                                                                               II.    RELATED  WORKThe model we present inthe next section is the result of inspiration we have taken from prior work onprediction on airline based on various factors. 18 This paper used theback-propagation neural network and genetic algorithm to forecast the airpassenger demand in Egypt(International and Domestic).

The factors thatinfluence air passenger are identified, evaluated and analyzed by applying theback-propagation neural network on the monthly data from 1970 to 2013. 19This paper is based on comparative study of new method and traditionalforecasting technique such as moving average, exponential smoothing,regression, etc. All methods were compared on the basis of a standard errormeasure- mean absolute percentage error. 20 This paper proposes an ensemble empirical mode decomposition(EEMD) based support vector machines (SVMs) modeling framework incorporating aslope-based method to restrain the end effect issue occurring during theshifting process of EEMD. Along with above research papers Oneof the most popular and frequently used stochastic time series models is theAutoregressive Integrated Moving Average (ARIMA) 3, 5 model. The basicassumption made to implement this model is that the considered time series islinear and follows a particular known statistical distribution, such as thenormal distribution. ARIMA model has subclasses of other models, such as theAutoregressive (AR) 6, 8, Moving Average (MA) 5 and Autoregressive MovingAverage (ARMA) 4, 5 models. Forseasonal time series forecasting, Box and Jenkins 8 had proposed a quitesuccessful variation of ARIMA model, viz.

the Seasonal ARIMA (SARIMA) 5, 8.The popularity of the ARIMA model is mainly due to its flexibility to representseveral varieties of time series with simplicity as well as the associatedBox-Jenkins methodology 3, 5 for optimal model building process. But the severelimitation of these models is the pre-assumed linear form of the associatedtime series which becomes inadequate in many practical situations.  III PROPOSED SCHEME This section illustrate theproposed scheme. To solve the problem of airlines passenger time series prediction  we have developed a model which consists of aSequential , 2 LSTM and one Dense Layer. Fig 1 shows the flow of data acrossthe various layers of the  model.

   LSTM  LAYER                                                                    SEQUENTIAL LAYER:It is used to create a sequential model. It acts as alinear stack of layers. All the other layers like LSTM and Dense are added toit to create a model.LSTM LAYER:Long Short Term Memory networks –usually just called “LSTMs” – are a special kind of RNN, capable of learninglong-term dependencies. They were introduced by Hochreiter & Schmidhuber(1997)21.LSTMs are explicitlydesigned to avoid the long-term dependency problem.All recurrent neuralnetworks have the form of a chain of repeating modules of neural network. Instandard RNNs, this repeating module will have a very simple structure, such asa single tanh layer.

                                              Fig:2LSTMs also have this chain likestructure, but the repeating module has a different structure. Instead ofhaving a single neural network layer, there are four, interacting in a veryspecial way.                                    Fig:3DENSE LAYER:Adense layer is simply a layer where each unit or neuron is connected to eachneuron in the next layer.                                       IV PROPOSED METHODOLOGYThesteps that we had taken to solve the problem at had were:1.       Inthe first step we imported the required modules which were namely numpy ,matplotlb , keras, math , pandas and sklearn .2.

       Inthe second step the dataset “international-airlines-passenger.csv” was loadedusing pandas read_csv() function.3.       Inthe third step the data was normalised using MinMaxscaler() function of sklearnmodule .

4.       Inthe fourth step the dataset was broken into two parts for one for training(67%)and the other for testing(33%) and the broken data chunks were reshaped so thatthey could be feed into the model.5.       Inthe fifth step we trained  and fitted ourmodel.6.

       Inthe sixth step we made prediction for the training and testing dataset.7.       Andfinally in the last step graphs were plotted for the predicted values.                                    V RESULTS Theresult obtained after working through the above algorithm or methodology can besummarised using a graph . The graph so plotted shows how accurately the abovemodel predicts the actual data points.

                                              Fig:4                             Thekey to read the graph shown in Fig 4 is present in the table 1.    COLOUR DATA POINTS BLUE ACTUAL DATA POINTS ORANGE PREDICTED PONTS OF TRAINING DATA GREEN PREDICTED POINTS OF TESTING  DATA                               Table:1                            VI CONCLUSIONFromthe graph shown in Fig 4 we can conclude that the model so used works quitewell for recreating the values that had been used in training but lacksefficiency in predicting the unseen data. But overall the model works fairlywell. Also there is a sudden jump that can be seen while shifting from trainingto testing data.

                         VII FUTURE WORKInfuture one might look at extending the model by incorporating convolutionlayers and try making the model even more deeper so that the efficiency of thenetwork improves and the sudden observed change also fades away while movingfrom training to testing data points.                          VIII REFERENCES1 T. Raicharoen, C. Lursinsap, P. Sanguanbhoki, “Applicationof critical support vectormachine to time series prediction”, Circuits and Systems,2003.

ISCAS ’03.Proceedings ofthe 2003International Symposium on Volume 5, 25-28 May, 2003, pages: V-741-V-7442 G.P.

Zhang, “A neural network ensemble method withjittered training data for timeseriesforecasting”, Information Sciences 177 (2007), pages: 5329–5346.3 G.P. Zhang, “Time series forecasting using a hybrid ARIMAand neural networkmodel”,Neurocomputing 50 (2003), pages: 159–175.

 4 John H. Cochrane, “Time Series for Macroeconomics andFinance”, Graduate Schoolof Business,University of Chicago, spring 1997.5 K.W. Hipel, A.I. McLeod, “Time Series Modelling of WaterResources andEnvironmentalSystems”, Amsterdam, Elsevier 1994.6 J.

Lee, “Univariate time series modeling and forecasting(Box-Jenkins Method)”, Econ413, lecture4.7 C. Hamzacebi, “Improving artificial neural networks’performance in seasonal timeseriesforecasting”, Information Sciences 178 (2008), pages: 4550-4559.8 G.E.P. Box, G. Jenkins, “Time Series Analysis,Forecasting and Control”, Holden-Day,San Francisco,CA, 1970.

9 H. Tong, “Threshold Models in Non-Linear Time SeriesAnalysis”, Springer-Verlag,New York,1983.10 Chu KL, Sahari KSM. Behavior recognition for humanoidrobots using long short-term memory. 2016;13(6):172988141666336.

11 Palangi H, Deng L, Shen YL, Gao JF, He XD, Chen JS, et al. DeepSentence Embedding Using Long Short-Term Memory Networks: Analysis andApplication to Information Retrieval. IEEE-ACM Trans Audio Speech Lang.2016;24(4):694–70712 Palangi H, Ward R, Deng L.

Distributed Compressive Sensing: A DeepLearning Approach. IEEE Transactions on Signal Processing. 2016;64(17):4504–18.

13 Yue J, Zhao W, Mao S, Liu H. Spectral—spatial classification ofhyperspectral images using deep convolutional neural networks. Remote SensingLetters. 2015;6(6):468–7714 Hinton G, Deng L, Yu D, Dahl GE, Mohamed A, Jaitly N, et al. DeepNeural Networks for Acoustic Modeling in Speech Recognition. IEEE SignalProcessing Magazine.

2012;29(6):82–97.15 Dahl GE, Yu D, Deng L,Acero A. Context-Dependent Pre-Trained Deep Neural Networks forLarge-Vocabulary Speech Recognition. IEEE Transactions on Audio Speech Processing.

2012;20(1):30–42.16 MesnilG, He X, Deng L, Bengio Y. Investigation of recurrent-neural-networkarchitectures and learning methods for spoken language understanding.Interspeech. 201317 Hochreiter S, Schmidhuber J.

Long Short-Term Memory. NeuralComputation. 1997;9(8):1735–80. pmid:937727618 International Journal of Computer Application (0975-8887) on “AirlinePassenger Forecasting in EGYPT (Domestic and International) by M.M.MohieEl-Din, M. S. Farag and A.

A. Abouzeid. Department of Mathematics, Faculty ofscience, Al-Azhar University Nasr City, Cairo 31884, Egypt19 Journal of Revenueand Pricing Management, Volume 1, Number 4 on “Neural network forecasting forairlines : A comparative analysis ” by Lawrence R. Weatherford, *Travis W.Gentry and Bogdan Wilamowski.20  “Forecasting Air Passenger Traffic by SupportVector Machines with Ensemble Empirical Mode Decomposition and Slope-BasedMethod” by Yukun Bao, Tao Xiong, and ZhongyiHu Department of Management Science and Information System,School of Management, Huazhong University of Science and Technology, Wuhan430074, China 21 In addition to the original authors, a lot of peoplecontributed to the modern LSTM. A non-comprehensive list is: Felix Gers, FredCummins, Santiago Fernandez, Justin Bayer, Daan Wierstra, Julian Togelius,Faustino Gomez, Matteo Gagliolo, and Alex Graves.



I'm Katy!

Would you like to get a custom essay? How about receiving a customized one?

Check it out