Detecting stock price manipulation is a cat-and-mouse game. Manipulators have constantly devised new techniques to avoid detection. The majority of the related work employed supervised learning techniques, which necessitated known manipulation patterns as examples for their models to recognize. To catch unknown and never-before-seen manipulation, we used unsupervised learning to train deep neural networks for detecting stock price manipulation in order to detect unknown and previously unseen manipulation. The models were trained to recognize normal trading behaviors that were expressed in a limit order book. Anomaly trading actions that did not follow to the learned patterns were identified as manipulated. The strength of our method is that it does not require prior knowledge about the characteristics of manipulation. As a result, it is best suited for detecting new or unknown types of manipulation. Two model architectures were evaluated: autoencoder (AE) and generative adversarial networks (GANs). They were put to the test on six prosecuted real manipulation cases from the Stock Exchange of Thailand (SET). With a low false-positive rate, both models could identify five out of six cases. For practical application of the models, a strategy called “MinManiMax” was also proposed to optimize the decision boundary.
This paper proposes a novel technique to identify black spots (prone-to-accident road locations) using street view images. The proposed technique is derived based on the hypothesis that the characteristics of the surroundings of the road have an effect on the safety level of a particular spot, and is the first black spot classification technique that is fully environment-aware. Assessing four street view images around each spot, a distance-aware pixel accumulation is developed to extract information about the objects surrounding the road from a semantically segmented image. The accumulated vectors are then used to train fully-connected neural networks to identify black spots. Performance evaluations are conducted with street view images in Thailand, which represent a challenging scenario of analyzing road characteristics in developing countries, with one of the highest road traffic fatality rates and limited historical accident records. Comparisons between our proposed technique and previously proposed techniques are also provided. Experiments show that our proposed technique succeeds in classifying black and safe spots in Thailand with an accuracy of 69.91%, where 75.86% of the black spots are identified correctly. Also, the distance-aware pixel accumulation can improve the accuracy of those machine learning techniques up to 6.4%. Our findings also evidently revealed that the object surrounding the roads as well as their sizes and distances are determinants of road’s accident proneness.
16 October 2017 in Journal of Advanced Transportation
This paper proposes an algorithm for real-time driver identification using the combination of unsupervised anomaly detection and neural networks. The proposed algorithm uses nonphysiological signals as input, namely, driving behavior signals from inertial sensors (e.g., accelerometers) and geolocation signals from GPS sensors. First anomaly detection is performed to assess if the current driver is whom he/she claims to be. If an anomaly is detected, the algorithm proceeds to find relevant features in the input signals and use neural networks to identify drivers. To assess the proposed algorithm, real-world data are collected from ten drivers who drive different vehicles on several routes in real-world traffic conditions. Driver identification is performed on each of the seven-second-long driving behavior signals and geolocation signals in a streaming manner. It is shown that the proposed algorithm can achieve relatively high accuracy and identify drivers within 13 seconds. The proposed algorithm also outperforms the previously proposed driver identification algorithms. Furthermore, to demonstrate how the proposed algorithm can be deployed in real-world applications, results from real-world data associated with each operation of the proposed algorithm are shown step-by-step.
Privacy policies, intended to provide information to individuals regarding how their personal data is processed, are often complex and challenging for users to understand. Businesses often demonstrate non-compliance with personal data protection laws, ranging from the absence of privacy policies to the existence of policies that do not adhere to legal requirements. This paper aims to (1) develop a quantitative and systematic tool for evaluating privacy policies' compliance with the Personal Data Protection Act (PDPA), (2) assess compliance among Small and Medium Enterprises (SMEs) in Thailand, and (3) provide recommendations for enhancing compliance practices. To achieve this, we proposed a multi-criteria privacy policy scoring model integrated with comprehensive statistical data analyses. The privacy policy scoring model consists of ten privacy principles and 31 privacy criteria, providing a structured framework for evaluating privacy policies. During a two-year postponement period for enforcing the PDPA law, we conducted a stratified random-sampling survey of 384 SMEs to evaluate their privacy policies using the proposed scoring model. The accomplished results revealed significantly lower scores than anticipated, with the nationwide average score of SMEs reaching only 6.1909 out of 100 points. More than half of the SMEs collected personal data without announcing privacy policies, and those with privacy policies adhered to an average of only 12.15 out of 31 privacy criteria. These findings highlight the pressing need to improve compliance practices among SMEs in Thailand. The proposed methodology can be customized and applied to align with the requirements of personal data protection laws in other countries. Additionally, our findings indicate that compliance with the PDPA is influenced by the Thailand Standard Industrial Classification (TSIC) sections, suggesting the adoption of tailored approaches by policymakers to address the specific needs of different TSIC sections.
1 June 2016 in International Journal of Trade, Economics and Finance