Predicting the number of customer transactions using stacked LSTM recurrent neural networks

Time series forecasting is used to predict future values on a data sequence using time-based data. Time series of transactions or count data are difficult to predict because of their complex nonlinear patterns. The purpose of this study is to provide a solution for predicting the number of upcoming transactions in order to predict the anomaly of the system, with the purpose of checking for why the number of transactions in the next day to be lower than expected. This can be a kind of customer churn prediction and can help system administrators identify and prevent potential losses and can improve existing infrastructure. Required data were collected from 780 company, which contains 353 rows of transactions in 2018 and has two variables, date and number of transactions; the number of transactions has been set as a target to predict. The autoregressive integrated moving average (ARIMA) model cannot deal with nonlinear relationships, while the neural network model, conversely, is capable of processing nonlinear patterns. In this study, a deep learning approach is presented. The approach is a stacked long short-term memory (LSTM) structure, a model derived from recurrent neural networks. To find the most accurate prediction model, the performance measures of various recurrent neural network, PROPHET, and ARIMA models are compared using the same dataset. Experimental results show that for predicting the number of transactions, the stacked LSTM model is better than other approaches.

[1]  Noah A. Smith,et al.  Greedy Transition-Based Dependency Parsing with Stack LSTMs , 2017, CL.

[2]  Yongjian Wang,et al.  DTDR-ALSTM: Extracting dynamic time-delays to reconstruct multivariate data for improving attention-based LSTM industrial time series prediction models , 2021, Knowl. Based Syst..

[3]  Hemanshu R. Pota,et al.  Fast Univariate Time Series Prediction of Solar Power for Real-Time Control of Energy Storage System , 2018, Forecasting.

[4]  Tianrui Li,et al.  Multivariate time series forecasting via attention-based encoder-decoder framework , 2020, Neurocomputing.

[5]  Dirk Neumann,et al.  Automated news reading: Stock price prediction based on financial news using context-capturing features , 2013, Decis. Support Syst..

[6]  Amir H. Mohammadi,et al.  Decline curve based models for predicting natural gas well performance , 2017 .

[7]  Jürgen Schmidhuber,et al.  LSTM: A Search Space Odyssey , 2015, IEEE Transactions on Neural Networks and Learning Systems.

[8]  Yuteng Xiao,et al.  A dual‐stage attention‐based Conv‐LSTM network for spatio‐temporal correlation and multivariate time series prediction , 2021, Int. J. Intell. Syst..

[9]  Rob J Hyndman,et al.  25 years of time series forecasting , 2006 .

[10]  Zhi-Ping Fan,et al.  Product sales forecasting using online reviews and historical sales data: A method combining the Bass model and sentiment analysis , 2017 .

[11]  Amy Loutfi,et al.  A review of unsupervised feature learning and deep learning for time-series modeling , 2014, Pattern Recognit. Lett..

[12]  Limin Xiao,et al.  Workload time series prediction in storage systems: a deep learning based approach , 2021, Cluster Computing.

[13]  Francis Eng Hock Tay,et al.  Support vector machine with adaptive parameters in financial time series forecasting , 2003, IEEE Trans. Neural Networks.

[14]  Pascal Vincent,et al.  Representation Learning: A Review and New Perspectives , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[15]  Durdu Ömer Faruk.  A hybrid neural network and ARIMA model for water quality time series prediction , 2010, Eng. Appl. Artif. Intell..

[16]  Nicolau Santos,et al.  Performance of state space and ARIMA models for consumer retail sales forecasting , 2015 .

[17]  Peter Stagge,et al.  Recurrent neural networks for time series classification , 2003, Neurocomputing.

[18]  Jianwen Guo,et al.  Mechanical fault time series prediction by using EFMSAE-LSTM neural network , 2020 .

[19]  Vinícius M. A. de Souza,et al.  Evaluation of statistical and machine learning models for time series prediction: Identifying the state-of-the-art and the best conditions for the use of each model , 2019, Inf. Sci..

[20]  Insup Lee,et al.  Stacked LSTM based deep recurrent neural network with kalman smoothing for blood glucose prediction , 2021, BMC Medical Informatics and Decision Making.

[21]  Yang-Ping Yao,et al.  Machine learning for pore-water pressure time-series prediction: Application of recurrent neural networks , 2021 .

[22]  Kyungyong Lee,et al.  Time-Series Analysis for Price Prediction of Opportunistic Cloud Computing Resources , 2018 .

[23]  Ravi Sankar,et al.  Time Series Prediction Using Support Vector Machines: A Survey , 2009, IEEE Computational Intelligence Magazine.

[24]  Min Xie,et al.  The use of ARIMA models for reliability forecasting and analysis , 1998 .

[25]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.