The latter feature that stands for the number of consecutive holidays ahead is one of the proposed attributes in this paper for the first time. In aiming to provide the feature matrix for machine learning models, some new influential variables are added to the literature. The improvement of forecasting ATMs cash demand of iran banking network using convolutional neural network. 13A, included data preprocessing, tuning hyperparameters, predictive regressors, and evaluating the performance of models. The authors hereby declare that they have no conflict of interest that would have affected the work presented in this article. Below is the link to the electronic supplementary material. A lot of machine learning models are available to choose from and deciding where to start can be intimidating. (ATM 1) Comparison of performance measures for different models (parametric: MA, SES, HES, ARIMA, and SARIMA; non-parametric-data-sequence: MLP_DS, SVM_DS, RF_DS, and KNN_DS; and non-parametric-regular-features: MLP, SVM, RF, and KNN) in the prediction of cash demands with approximate iteration. Due to vacillating users demands and seasonal patterns, it is a very challenging problem for the financial institutions to keep the optimal amount of cash for each ATM. 11, non-parametric models can reliably predict the changes in the withdrawal pattern both before and during COVID-19. In the first algorithm, a named data-sequencethe cash withdrawal of the preceding days (i.e., prior 7days)is used as the input variables to forecast any new observation's cash withdrawal data. 1Department of Information Systems and Analytics, Farmer School of Business, Miami University, Oxford, OH 45056 USA, 2Department of Chemical Biomolecular Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH 45701 USA, 3Department of Industrial and Systems Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH 45701 USA. ATM Cash Prediction Using Time SeriesApproach, in . Are artificial neural networks black boxes. Also, the chronological cash New Dataset. improved accuracy of the cash demand forecasts due to reduction in computational complexity when predicting an ATMs daily cash demand for groups of ATM centers with similar day-of-the week cash withdrawal seasonality patterns. Motivated by a real-world case in a private bank, this paper precisely addresses this gap in the literature by proposing an extensive evaluation that can forecast ATM cash demand before and during the COVID-19 outbreak. The structure of the network (number of layers: n-hidden-layers and number of neurons in each layer: n-nodes) is the critical part of this model required to obtain the highest possible accuracy. Bao Y, Xiong T, Hu Z. Multi-step-ahead time series prediction using multiple-output support vector regression. New Notebook. By definition, stationary time series hold the p value and test statistics lower than 0.05 and critical values, respectively (see Table Table22 for more details). 2, ATMs 1 and 2 have a high cash demand during weekdays, followed by a low amount of money withdrawn on weekends. Parametric techniques are statistical methods that require some prior knowledge about the distribution of data [33]. Simutis R, Dilijonas D, Bastina L, Friman J. ATM (a group of ATMs can also be worked that is treated as a single ATM) to develop a Comparison of ATMs in terms of MSE, POCID, and Fitness metrics, Details of performance measures for each model are reported in Tables S5S14. The following code snippet separates the numeric features, selects the categorical features and use one-hot encode on these features, and joins the two sets together. LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, Are you sure you want to create this branch? (Colors codegreen: best parametric model, purple: best non-parametric-data-sequence model, and yellow: best non-parametric-regular-feature model.) Learn more about the CLI. Similarly, Fig. of this software and associated documentation files (the "Software"), to deal Brownlee J. Careers, Unable to load your collection due to an error. 12A) the RF learning predictor has the highest performance both in terms of prediction error and the accuracy of directions changes, while during COVID-19 (Fig. Moreover, the demand for cash is not only influenced by time, but it follows different tendencies that make modeling even more difficult. The actual withdrawals (total_amount_withdrawn) includes all the transactions where the actual amount is withdrawn from the ATM. The .gov means its official. Such a significant prediction can help bank managers to mobilize idle cash and generate additional revenuerather than load excess banknotes in ATMs, which increases operational and opportunity costs, especially when there are thousands of ATMs. Cash Demand Forecasting of ATMs. [12] presented a model based on time series and regression using the 3-year data from a bank in Serbia. The MSE is computed to figure out how far the prediction values are from the actual values in terms of quantity. Finally, SectionConclusion reports the conclusion and possible directions for future work. We use cookies on Kaggle to deliver our services, analyze web traffic, and improve your experience on the site. Additionally, the maximum mean squared errors before and during COVID-19 are about 1% and 5%, respectively. The SVM has been successfully applied to a variety of different classification and regression problems [2, 25, 36]. A shows all available data, while B magnifies the last 3 months of the dataset for the sake of clarity. intelliCast can save you over 20% in the dead cash using our proven forecasting techniques. HHS Vulnerability Disclosure, Help Therefore, in this This test reports the p value and test statistics of the given time series. In contrast, non-parametric approaches, known as machine learning (ML) prediction methods, have no such limitations and are ready to be applied to any non-linear series, as the distribution of data is not an issue [28, 33]. X = features.copy().drop(columns = ['total_amount_withdrawn', 'trans_date_set', 'trans_month','trans_year', 'working_day_H', 'working_day_W']), Xtrain, Xtest, ytrain, ytest = train_test_split(X, y, shuffle= False, test_size = 0.2, random_state = 42), shap_values = shap.TreeExplainer(xgb).shap_values(Xtest), # Transform categorical features into the appropriate type that is expected by LightGBM, # Splitting the dataset into the Training set and Test set. Also, the chronological cash demand for every ATM fluctuates with time and often superimposed with non-stationary behavior of users. Barrow D, Kourentzes N, Sandberg R, Niklewski J. The prediction of the change in direction (POCID) metric, denoted by Eqs. However, in case there is normal volatility in the time series, these features play a more pivotal role than the preceding days cash withdrawal information. According to the results represented in Figs. In: Springer series in statistics. One of the well-known strategies in multi-step time series forecasting is a recursive strategy (also known as walk-forward). Qiu X, Zhang L, Suganthan PN, Amaratunga GA. Oblique random forest ensemble via Least Square Estimation for time series forecasting. Venkatesh et al. This result shows that category-wise forecasting based on the accessibility, environment factors, and different withdrawal patternsrather than taking the average of daily cash demand from different ATMssignificantly enhances the prediction quality. We will work on the demand for a single ATM (a group of ATMs can also be worked that is treated as a single ATM) to develop a model for the given data set. Banks normally pay a significant amount of fixed fees for the re-filling, additional extra cost for the transportation with security arrangement. The greater Fitness value indicates better forecasting of fluctuations in time series and better accuracy of the prediction model. To compensate the ARIMA limitation for the series with seasonality, the SARIMA model plays a prominent role by taking the seasonal autoregressive order (P), seasonal difference order (D), seasonal moving average order (Q), and the number of time-steps for a single seasonal period (s) into account [33]. To visually compare the performance of the three categories of employed models (i.e., parametric, non-parametric with data-sequence, and regular-features) before and during COVID-19, the prediction of the best predictive model of each category versus the actual data are plotted in Figs. Federal government websites often end in .gov or .mil. This paper strived to conduct a trustful comparison approach and perform different models equally well. Even LP can be used solving multiple objectives at a time. For example, how holidays affect the use of ATM depends on where the teller is located. Automatic robust estimation for exponential smoothing: Perspectives from statistics and machine learning. The algorithms were performed on a computer with a Windows 10 operating system and a CPU Core i7-6700 with 16GB of memory. and also to which of the similarly anonymous cities it belongs. Lima Junior AR. We have to keep in mind that, the performance of the trained model deteriorates as the size of the training data set shrinks. This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository. For both strategies, a history of feature matrix and target vectors have been constructed to generate the future observations (next days) input variables; however, to prevent the forecasted observations from entering into the learning process, this history was not added to the training set when the model was fitted. Choose save to set your Quick cash preference. Results obtained with the approximate iteration strategy in Fig. However, this metrics limitation is that the prediction trend in time series is not clearspecifically when the pattern abruptly changes. The non-parametric models employed in this study are well-known ML regressors, namely MLP, SVM, RF, and KNN. Aghaaminiha M, Mehrani R, Reza T, Sharma S. Comparison of machine learning methodologies for predicting kinetics of hydrothermal carbonization of selective biomass. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR S1B, C for more comparison of ATMs in terms of time-related features. The maximum mean squared errors before and during COVID-19 were about 1% and 5%, respectively. However, the last month of the third year has a contrary trend because of the beginning of the COVID-19 pandemic and the announcement of a stay-at-home order. In previous studies in this context, different research papers came up with varying conclusions regarding the performance of machine learning models (non-parametric predictors) compared to the statistical models (parametric predictive models). The models suffixed _DS constitute ML models in which a data-sequence algorithm was used to build their feature matrix. There could be seasonality, e.g. This sounds like complex work, but it is relatively simple and straightforward. Before the COVID-19pandemicin times when there were only minor disturbances in withdrawal patternsforecasting quality was higher, and generally, the non-parametric models could more accurately predict the ATMs money demand. Another well-known ML regressor is RF, which constructs a combination of multiple decision trees for regression purposes [34]. S12S14 to see the results of other ATMs. For instance, the results of this study were discussed based on short-term forecasting horizon during pandemic (30days period for before and during COVID-19) to determine the most appropriate models in forecasting cash demand. For example, Adebiyi et al. Thus, the results of approximate and updated iterations are the same for these models. However, using a more meticulous outlook, the nearby-location and/or the same geographical location of ATMs do not necessarily indicate a similar withdrawal pattern, since the points of interest in the ATMs vicinity might be different. Besides, only a few papers considered both time- and location-related variables (e.g., [20, 21]), though the location of ATMs can meaningfully affect the amount of daily cash withdrawn from these ATMs. learning models to troubleshoot the above use case. [43] stated that chaos was present in the NN5 Competition dataset; therefore, to appropriately estimate the cash demand, they used the TISEAN tool to calculate the optimal lag and embedding dimension of each series. the contents by NLM or the National Institutes of Health. In this study, it has been revealed that error measure (e.g., MSE) alone cannot be the best evaluation metric in comparing the performance of the predictors on ATM cash demandespecially when the withdrawal pattern drastically changes as a result of preventive measures such as a stay-at-home order or partial lockdowns that are taken to reduce the spread of COVID-19. MSACD Spring'18 competition - ATM CashFlow Prediction. Applying the predictive models on diverse (not a single) time series datasets can reduce the raising of questions regarding the statistical significance of our results and generalization [31]. We can see from the histogram that, the center is near 60. The ATM demand forecasting problem became more popular after the Forecasting Competition for Artificial Neural Networks and Computational Intelligence (NN5 Competition) in [17]. It is always advisable to conduct an extensive mining work to extract hidden information from the data. Taking comprehensive hyperparameter tuning into account, the reasons for such impressive results might be mainly related to the high performance of ARIMA and SARIMA for short-term prediction [44] and the fact that we aimed to predict the demand just after the occurrence of the pandemic, while avoiding or minimizing overfitting. Although this paper addressed some gaps in the literature, some limitations still need to be tackled in the future to further enhance the forecasting performance. In the case of Iran, for example, the number of ATMs and debit cards are about 60 thousand and 23 million, respectively [14]. Figure1 illustrates the time series of cash withdrawals from ATMs 1, 2, and 3, while Fig. Lim B, Zohren S. Time series forecasting with deep learning: a survey. 9A with Fig. OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE demand for every ATM fluctuates with time and often superimposed with non-stationary Generally, ATM means that contains the average daily cash withdrawal for all three ATMs has a higher MSE, as well as a lower POCID and Fitness, in all associated configurations, resulting in lower overall performance. Therefore, developing cash demand forecasting model for ATM network is a challenging task. Having historical . Zandevakili M, Javanmard M. Using fuzzy logic (type II) in the intelligent ATMs cash management. The SES model behaves similarly, but unlike MA, it uses exponentially decreasing weights for past datapoints. table_chart. In machine learning exercises, there are three broader parts: (1) data extraction & mining which helps to decide on the features (this normally takes around 6070%), (2) decide and fit a model which includes hyper-parameter optimization(this normally takes 1015%), (3) accuracy metrics & testing takes 1015% of time). However, some significant features, such as the number of consecutive holidays ahead, have not been included in the previous studies. The datasets are publicly available at https://github.com/af551515/Forecasting_ATM_Cash_Demand. The higher value of POCID, the better the mapping of the trend. Note that the special days are Mothers Day, Fathers Day, Teachers Day, Students Day, College Day, Valentines Day, Love Day, and Yalda (national day). However, the exact types of special days are not considered; instead, yes or no values are employed. We have daily transactions data from 2011 till 2017. Figures S2S4 contains the same information for other ATMs. J Appl Math. Thus, the pandemic condition is perfectly observed and captured with SARIMA, and it can better map further withdrawal patterns from ATMs during COVID-19. Business would probably be interested to see a final tabular report. forecasting model for ATM network is a challenging task. Finally, a modified fitness measure is proposed for the first time to correctly choose the most promising model by considering both the prediction errors and accuracy of directions change simultaneously. A methodology to improve cash demand forecasting for ATM network. and transmitted securely. XGBoost is a . Notebook. Taieb SB, Bontempi G, Atiya AF, Sorjamaa A. Hierarchy of employed time series prediction models in this study, In the data-sequence algorithm, the feature matrix is constructed via the transposition of the data-sequence with the sliding window of length 7 (the yellow-shaded rows in Fig. The best predictive model in each class is reported with different colors (i.e., the parametric model in green, non-parametric-data-sequence in purple, and non-parametric-regular-features in orange). However, loading excess cash in ATMs, rather than only loading in what the demand roughly is, will increase operational and opportunity costs [9, 20, 21]. ATMs in each category have a similar distribution, and ATMs used in this study are the most representative of their group. However, the cash demand from ATM 2 (located in business districts) is lower because fewer peoplemostly personnel of the companies/agencies in the vicinityhave access to such ATMs. XGBoost model has done fairly well compared to LinReg and using default parameters with average accruracy of over 90% compared to approx 50% of LinReg . S15S17 to see the results of other ATMs, (ATM 1) Comparison of different ML methods to predict cash demand with updated iteration strategy. [39] obtained the best results with a multi-input, multi-output forecasting strategy that selected autocorrelation selection criteria using input variable selection, deseasonalization, and average weight combination. Variables importance plot is shown below: LightGBM offers good accuracy with integer-encoded categorical features. Please Ding S, Li Y, Wu D, Zhang Y, Yang S. Time-aware cloud service recommendation using similarity-enhanced collaborative filtering and ARIMA model. Alongside the contributions already mentioned, this study aims to fill the gap and propose a comprehensive evaluation for ATM cash demand prediction both before and during the COVID-19 pandemic (i.e., just after a disruption in demand) to choose the most promising algorithms based on a new performance metric that simultaneously takes both the . Alongside the contributions already mentioned, this study aims to fill the gap and propose a comprehensive evaluation for ATM cash demand prediction both before and during the COVID-19 pandemic (i.e., just after a disruption in demand) to choose the most promising algorithms based on a new performance metric that simultaneously takes both the error and accuracy of directions change into account. Inclusion in an NLM database does not imply endorsement of, or agreement with, It was revealed that during COVID-19, in which there was a sudden shock in demand followed by abnormal volatility in withdrawal patterns, the parametric models of ARIMA and SARIMA could mostly provide better predictions based on the Fitness evaluation metric. The same splitting approach was used for the other ATMs (see Figs. Importance of tuning hyperparameters of machine learning algorithms. In this model, the number of lags observed in the data (p), the number of times datapoints are subtracted to make the series stationary (d), and the size of the moving average window (q) are the substantial hyperparameters that should be carefully chosen based on some statistical analysis [27, 33]. NN5_FINAL_DATASET_01. With that aim, the models were implemented and compared after performing an exhaustive statistical analysis, coupled with grid search and k-fold cross-validation techniques that led to the highest performance of models. With that aim, we first identify three different categories for ATMs based on their accessibility and environmental factors, which significantly affect both the daily cash demand and the withdrawal pattern. By using Kaggle, you agree to our use of cookies. To obtain a decent range for other parameters of the non-parametric models, we estimated the ACF and PACF plots of cash withdrawal data for both d values of 0 and 1 (Fig. However, cash demand is inherently comes with high variance and non-stationary stochastic process which can affect the reliability of many approaches. S8). According to the results, before COVID-19, non-parametric models outperform the parametric methods in all eight configurations. The same approach was employed for the categories of holidays that are New Year (4days), Religious Holidays (17days), and National Holidays (6days)., (ATM 1) The importance of employed features in non-parametric-regular-feature models. One primary assumption of models in the literature is that the amount of cash demand and withdrawal patterns are not overly volatile (though some studies have investigated chaos time series and uncertainty in demand). The rest of this paper can be summarized as follows. For the tuning process, a range of values for the hyperparameters of all models is considered, and accordingly, several sub-models are then constructed. IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, MA: conceptualization, methodology, software, formal analysis, writingoriginal draft, and writingreview & editing. 70.1s - GPU P100. S2S4). The architecture of built A parametric models, B non-parametric-data-sequence models, and C non-parametric-regular-features models. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Work fast with our official CLI. [4] were ranked first among the computational intelligence models in the competition. The former utilized self-organizing fuzzy neural networks and obtained 21.5% SMAPE, while the latter predicted time series with recurring seasonal periods and developed a model based on a combination of forecasting methods via a simple average of forecasts, achieving 22.2% SMAPE. The former obtained a Fitness at 68.87, and the latter achieved 71.57. Ekinci Y, Lu J-C, Duman E. Optimization of ATM cash replenishment with group-demand forecasts. The simple exponential smoothing model. Financial institutions (e.g., banks, credit unions, and stock brokerages) might have thousands of ATMs and, in turn, millions of transactions over the course of a year. IEEE, 2014; p. 15. Because of this, over the past years, the number of ATMs in the world has increased, reaching over 3 million machines [7]. sign in In this study, the performance of predictive models on forecasting the cash demand for different ATMs both before and during the COVID-19 pandemic with approximate and updated iteration strategies are extensively evaluated. If the forecast is wrong, it induces a considerable amount of costs. Ramrez C, Acua G. Forecasting cash demand in ATM using neural networks and least square support vector machine. SARIMA represented the highest performance among all predictive models owing to its closest prediction to the actual data (small prediction error). Simutis et al. Our average prediction is quite accurate here. The competition had some rules and assumptions, such as using the same model for all ATMs; using the same performance measure of symmetric MAPE (SMAPE); 2-year time frame data of 111 empirical ATM daily cash withdrawal series in England; and a demand prediction for the next 56days. The number of trees (n-trees) and the fraction of features used to grow each tree (max-features) are the primary hyperparameters that need to be tuned for this method [29]. In this contexta sudden shock in demand and massive volatility in the withdrawal patternselecting the most efficient model with appropriate diagnostic performance is of paramount importance as it relates to accurately predicting ATM cash demand. 2020. arXiv: 200707588 [cs, stat]. Adebiyi AA, Adewumi AO, Ayo CK. Some other works for NN5 Competition data were presented by Coyle et al. In: The 2010 International Joint Conference on Neural Networks (IJCNN). Moreover, none of the previous ATM cash withdrawal time series contains a huge amount of volatility that stemmed from a disaster or unprecedented challenge (e.g., pandemic). It is quite obvious that daily cash withdrawal amounts are time series. See Fig. Additionally, category-wise forecasting led to improving forecasting quality by at least 4%. SA: methodology, software, formal analysis, and writingoriginal draft. Predicting closed price time series data using ARIMA Model. Before COVID-19 (Fig. between 710th day of each month some people get their pension. Abstract. In the literature, many studies examined time-related independent variables to capture the seasonality in the data. Bat , Gzpek D. Joint optimization of cash management and routing for new-generation automated teller machine networks. Bethesda, MD 20894, Web Policies In general, there are two types of time series prediction approaches, namely, (i) parametric and (ii) non-parametric. The ARIMA model makes its prediction using the difference between the values of datapoints, rather than their actual values. However, choosing the most efficient model to appropriately forecast an ATMs cash demand is one of the most important activities. Wichard JD. Arora N, Saini JKR. Looking at the ACF, a seasonal lag of 7, 14, 21, etc. Thus, a more accurate prediction of ATM currency demand can help financial institutions avoid being tempted to fill ATMs with too many notes and earn more profit by mobilizing idle cash and generating additional revenue through investmentsspecifically in countries with high-interest rates and overnight interest rates. Figure8 depicts the results and produces the rank of features in such a way features, including day of year, day of month, weekday, month, and n HDs ahead are the most influential attributes selected for ATM 1.
Credit Suisse Fintech Conference,
Esp32 Module Datasheet,
Fivetran Revenue 2022,
Articles A