Time Series and Neural Network Analysis

A review of artificial intelligence quality in forecasting asset prices

“…The stocks of Apple, IBM, and Microsoft were the most commonly studied due to previous work being readily accessible as a benchmark. The best results for Apple's stock price were obtained by Araújo (2010), Asadi (2019), and Tseng et al (2012), while Borovkova and Tsiamas (2019), Hadavandi et al (2010b), Lahmiri (2018), and Rout et al (2014) had the best performance for IBM's stock price data, and González‐Mancha et al (2017), Kazem et al (2013), and Sheelapriya and Murugesan (2017) for MSFT.…”

“…Conversely, Cao and Tay (2001) applied the relative difference of the value at time t to t À 5, stating that this method had four advantages, with the most significant being that it made the data distribution closer to normal, thereby improving its predictive capacity. Tseng et al (2012) conducted a study dedicated to investigating the benefits of using standardized data. Using data from 50 different companies' stock assets, the authors built and tested prediction models using five different formats, including ANNs with normalized data and a network with raw data without normalization.…”

“…The sample size was sufficient for the model to achieve a MAE of 0.116 in normalized data, compared to 0.145 of the best model in the benchmark (Näive). It is worth noting that studies on HFT data are relatively recent, but they have gained significance in both practical applications (Tseng et al, 2012) and academia.…”

“…Table 4 provides a descriptive summary of the data scaling, feature selection, and hybrid model categories identified in the sample of articles each year, with responses indicated as either “yes” or “no.” The use of preprocessing techniques is quite common, particularly for neural networks (Qi, 2002; Tseng et al, 2012), and there are numerous ways to do so. The most widely used techniques include normalization using min‐max (Bhowmick et al, 2019; Das & Padhy, 2018; Das et al, 2015; Hajibabaei et al, 2014; Hegde et al, 2018; Shrivastava & Panigrahi, 2011) or normal distribution approximation (Van Gestel et al, 2006; Yazdani‐Chamzini et al, 2012), also referred to as standardization (McNally et al, 2018), as well as softmax function (Jiang et al, 2018), log‐transform (Gradojevic & Yang, 2006; Kumar, 2010; Kumar & Thenmozhi, 2012; Matyjaszek et al, 2019), differences (Cao & Tay, 2001; Gradojevic & Yang, 2006), and those articles that do not specify any data transformation.…”

“…The forecasting process focuses on prices and related aspects, such as return (Bollen et al, 2011;Jasic & Wood, 2004;Kumar, 2010;Kumar & Thenmozhi, 2014a, 2014bMatías & Reboredo, 2012;Sermpinis et al, 2016;Stasinakis et al, 2016;Tseng et al, 2012) and volatility (Das & Padhy, 2018). In financial markets, prices from various sources are available, including stocks, indices, commodities, foreign exchange, and bonds, which are the primary data and measures of interest for forecasting models.…”