Abstract: Cloud service providers use AI capabilities to provide clients with AI Service in addition to usual cloud services today. In such a scenario, business departments can use these AI Services conveniently by just calling APIs. For example, the price forecasting can be done by inputting time series data into an AI Service without having the Data Scientists or ML Engineers to build model training or hyper-parameter tuning. However, AI models with complex structures are difficult to interpret and explain as black boxes. As a result, it is essential to monitor the distribution of input and prediction results and show alert signals to end-users as well as business departments. As model governance is crucial for AI compliance, on the one hand, there come questions about how to monitor cloud-based AI services to enhance interpretability and transparency. On the other hand, to analyze time series-based data and gain insightful views, auto-ML technology is becoming popular to build, analyze, and optimize time-series forecasting models. In addition to black-box models, such custom-built model deployment applications also need built-in rate sampling, distribution, and prediction monitoring mechanisms after deployment. Furthermore, there exists a gap in the visual exploration of time-series forecasting model analysis and monitoring in contrast to the explainability of the prediction as an algorithm-agnostic solution. Many methods have been proposed to deal with time-series data, and how to provide an effective and efficient data API is essential. In the big data era, modern applications generate massive volumes of time-series data as a result of the high frequency of measurements from IoT devices, sensors, and financial transactions. Data mountains cause problems in solving data storage, computation, and analytics, and there is a timely review of big time-series data, a class of big data. With the increased need for data-in-cloud pattern recognition and intelligence discovery, temporal data mining has attracted growing attention. A lot of pattern mining methods have been proposed, and in contrast, visualization support on temporal data mining is scarce.
Keywords: Wealth Management,Artificial Intelligence (AI),Financial Services,Cloud Computing,Data Engineering,Next-Generation Finance,Robo-Advisory,Predictive Analytics,Portfolio Optimization,Digital Transformation,Fintech,Big Data,Machine Learning,Intelligent Automation,Customer Personalization.
|
DOI:
10.17148/IJARCCE.2022.111250
