Welcome! This is the documentation section for SenseMatrix's ATM analytics product: PlutoXATM.

Introduction

PlutoXATM is an artificial intelligence product by SenseMatrix. It consumes ATM data to learn the hidden patterns behind the chaos transactions. At the same time PlutoXATM forecasts the cash flow of deposits and withdrawals in a real-time manner.

It is important that banks and finance organizations can predict their ATM cash flow. Accurate forecasts can greatly improve the experience of business operations; reduce the cost of process; enhance financial plan; increase the efficiency of using cash flow. PlutoXATM provides a learning knowledge based infrastructure that can support financial, security, and visualization applications for many aspects. Some examples are:

  • - Predict reload ATM cash time points and prepare the cash amounts. An accurate cash flow estimate can result in a better financial plan. It reduces the operation costs and increases the efficiency of using cash flow.

  • - Forecast the short-term and long-term trends. It potentially identifies the customer behaviors related with geolocation, time period, customer segmentation, holidays, or other hidden perimeters.

  • - Detect financial frauds and anomaly activities. By predicting and monitoring the cash flow, it provides such abilities to detect the financial frauds and software/hardware failures.

    • - Visualize ATM performance and efficiency. It gives better understanding of ATM usages, which could be used for improving the ATM layout and cash flow plan or design.

PlutoXATM lays on a real-time streaming framework such as Apache Kafka. PlutoXATM can be viewed as a data stream AI engine, which digests continuous transactions and outputs the learned patterns and predictions in a real-time manner. PlutoXATM includes four major components. They are:

  • Data Interface : PlutoX Data Interface stands between the boundary between PlutoXATM kernel and the data stream framework such as Kafka. Interface not only digests the messages from the lower layer, but also pushes the prediction back to a stream pipeline for other application consuming.

  • DataLake : PlutoX DataLake is a standalone component within PlutoXATM. It responses data cleaning, enrichment, and normalization. It also buffers and shuffles the chaos data and maintains the data consistence/integrity to guarantee the accuracy of training and prediction.

  • Pattern Learning : PlutoX Pattern Learning learns the knowledge from the transactions. The learned knowledge is viewed as the signatures or patterns. These patterns may be related with one or combined multiple parameters such as geolocation, time period, customer segmentation,weather, holiday, economy, etc. Some of these parameters are short-term related such as bad weather or software/hardware failures. Some impact the patterns in a long-term way such as economy. Pattern learning module is the most important component that works on a stream mining algorithm trained by all these parameters.

    • Prediction : PlutoX Prediction forecasts the future cash flows based on the trained patterns and real-time transactions. PlutoXATM prediction is made by considering both short-term and long-term parameters. Examples of short-term parameters include features such as weather, system failures, or holidays; long-term parameters consider the features such as customer behaviors, geolocation, or long period economical wavelets.

Why PlutoXATM?

There are multiple reasons that you might need to consider PlutoXATM as your solution.

  • Advanced Learning Algorithm : PlutoXATM is an artificial intelligence application to predict the cash flow of ATMs. It is designed based on an advance data stream mining algorithm. The prediction is made based on the trained patterns which are the knowledge learned from the huge amount of chaos transactions. Distinguished from these common statistic methods, PlutoX stream learning algorithm does not request of any predefined knowledge or assumption. Algorithm is designed to find the hidden patterns covering multiple learning parameters such as geolocation, time period, customer segmentation, weather, holiday and economy wavelets without any statistic assumptions.

  • Large Volume : PlutoXATM is designed as a big data stream learning/prediction engine. It supports up to 1.0E5 transactions per second. DataLake is designed in a parallel computation manner. Learning and prediction components are implemented with high efficiency for both time and space complexity manners. After tested, one instance of PlutoXATM can support no less than 1.0E5 ATMs and it can be easily expended by deploying multiple instance of PlutoXATM with segment of ATM zones.

  • Better User Experience : PlutoXATM learns patterns without pre-knowledge or pre-assumption. PlutoXATM training and prediction are designed as a one-time pass processing manner. In other words, transactions are loaded one and only one times for both learning and prediction purposes. Traditionally preparing a huge amount of historical dataset for train is a hard task. PlutoXATM provides two training options. PlutoXATM accepts historical dataset training. It helps the training converging in a fast pace. PlutoXATM also provides none pre-training option, which can be viewed as a Plug-and-Play mode. It avoids the tasks to prepare the historical data.

  • Open Architecture : PlutoXATM can either be an ultimate application, meaning that customer directly consumes prediction; or a middleware. It support other applications such as cash flow alerting, visualization, financial fraud detection, ATM software/hardware failure monitoring, anomaly detection, etc.

  • Saving Financial and Operational Costs : Reducing the financial and operational costs are the major reasons that banks want to predict the ATM cash flow. Accurate real-time forecasts help to reduce the costs in many aspects, which may include: (i) increase the efficiency of utilizing funds by having better forecasts; (ii) prevent and detect financial frauds in a realtime manner; (iii) monitor and detect real-time ATM software/hardware failures; (iv) visualize and predict customer behaviors in multiple dimensions such as geolocation and time periods.

Conclusion

PlutoXATM is an artificial intelligence module that learns the patterns and predicts the ATM cash flows. PlutoXATM is designed as plug-&-play to maximize the user experience. Learning and prediction are designed to consider many parameters such as ATM geolocation, software/hardware failures, time/date, weather, holiday and festivals, and other anomalous. Therefore, PlutoXATM can reach a high prediction accuracy by adjusting its prediction based on real time events accordingly. PlutoXATM prediction supports multiple business use cases such as cash flow planning, anomaly detection, and security alerting, etc. All of these use cases can bring customer the more core values of business in variety aspects. If you are wondering what you can expect from predicted results, please check out our demo site.