Abstract
As many automated algorithms find their way into the IT systems of the banking sector, having a way to validate and interpret the results from these algorithms can lead to a substantial reduction in the risks associated with automation. Usually, validating these pricing mechanisms requires human resources to manually analyze and validate large quantities of data. There is a lack of effective methods that analyze the time series and understand if what is currently happening is plausible based on previous data, without information about the variables used to calculate the price of the asset. This paper describes an implementation of a process that allows us to validate many data points automatically. We explore the K-Nearest Neighbors algorithm to find coincident patterns in financial time series, allowing us to detect anomalies, outliers, and data points that do not follow normal behavior. This system allows quicker detection of defective calculations that would otherwise result in the incorrect pricing of financial assets. Furthermore, our method does not require knowledge about the variables used to calculate the time series being analyzed. Our proposal uses pattern matching and can validate more than 58% of instances, substantially improving human risk analysts' efficiency. The proposal is completely transparent, allowing analysts to understand how the algorithm made its decision, increasing the trustworthiness of the method.

Abstract
Taxi demand prediction in a city is a highly demanded smart city research application for better traffic strategies formulation. It is essential for the interest of the commuters and the taxi companies both to have an accurate measure of taxi demands at different regions of a city and at varying time intervals. This reduces the cost of resources, efforts and meets the customers' satisfaction at its best. Modern predictive models have shown the potency of Deep Neural Networks (DNN) in this domain over any traditional, statistical, or Tensor-Based predictive models in terms of accuracy. The recent DNN models using leading technologies like Convolution Neural Networks (CNN), Graph Convolution Networks (GCN), ConvLSTM, etc. are not able to efficiently capture the existing spatio-temporal characteristics in taxi demand time-series. The feature aggregation techniques in these models lack channeling and uniqueness causing less distinctive but overlapping feature space which results in a compromised prediction performance having high error propagation possibility. The present work introduces Spatio-Temporal Aggregator Predictor (ST-A(G)P), a DNN model which aggregates spatio-temporal features into (1) non-redundant and (2) highly distinctive feature space and in turn helps (3) reduce noise propagation for a high performing multi-step predictive model. The proposed model integrates the effective feature engineering techniques of machine learning approach with the non-linear capability of a DNN model. Consequently, the proposed model is able to use only the informative features responsible for the objective task with reduce noise propagation. Unlike, existing DNN models, ST-A(G)P is able to induce these qualities of feature aggregation without the use of Multi-Task Learning (MTL) approach or any additional supervised attention that existing models need for their notable performance. A considerable high-performance gain of 25 - 37% on two real-world city taxi datasets by ST-A(G)P over the state-of-art models on standard benchmark metrics establishes the efficacy of the proposed model over the existing ones.

Abstract

Abstract

Abstract

Abstract