HTF Chart Aggregation

Aggregated Chart Data built for predictive models.

This feature has been in development, but Nosam (the model) with the target of active trading is halted. Nosam is now being used in AI-HFTs TE aggregation.

Chart data is perfectly designed for neural networks. Candlestick charts consist of a sequence of interval-based highs and lows. All efforts to train a neural network for technical analysis failed for one simple reason: the AI was trained purely on the data sequence of a specific time interval. Let's take a one-minute time interval per new data transmission as an example. Here, AIs only received the data of the opening price, the closing price, the high, the low and the volume within the one-minute time interval. ShezheaNET NosamAI is trained here in a highly differentiated way. Multiple time intervals are used for NosamAI: 1 second, 2 seconds, 5 seconds, 16 seconds, 26 seconds, 31 seconds, 1 minute, 1 hour. The AI is primarily fed with the same data transmissions as traditional AIs in these multiple time intervals. However, the general framework here is highly differentiated. Nosam is strongly based on liquidity and volume levels. Highly liquid price levels as well as high volume levels, such as standardised S&R levels, and institutional order block levels are strongly considered. Because NosamAI is trained on low time intervals, it can more easily recognise institutional order blocks, for example. Here, not only the highly efficient market levels are taken into account and listed, but also the particularly inefficient ones. In general, the framework is based on the ICT 2022 Entry Model by Michael Huddleston. A fixed entry framework is of high importance here, as the AI has a constantly similar environment in which it can operate. Nosam is trained purely on efficient markets. Here, the following markets are used as training environments: NQ, ES, DAX, EURUSD.

Modern security valuation and ratio analysis is time-consuming and expensive. ShezheaNET argues that the entire analysis of trends, individual stocks, and macroeconomic values is AI-driven, more efficient, less expensive, and more profitable. Social sentiment is best measured by real people, real unfiltered opinions from diverse groups of people. Hedgefund's research papers address macroeconomic as well as niche-specific trends from a purely mathematical objective viewpoint. This mathematical objective view, whose core approach is also important for ShezheaNET, is unreliable and says little about the sustainability of many trends. While current numbers give information about past quarters and most forecasts are based on old results, social sentiment for both macroeconomic and especially individual stocks gives a picture of the sustainability of a trend and the future of a particular sector or stock. Social sentiment makes it possible to "frontrun" certain sectors. That is, committing to a trend direction based on data not yet priced in before quarterly reports validate these trend direction decisions. Social sentiment can also be measured with data. A lot of publicly readable data, such as monthly users or visitors to a website, can ultimately be crucial to the final outcome in the quarterly report. Other data that reports on social sentiment are fast-paced social media algorithms such as TikTok and X (formerly: Twitter). TikTok is heavily weighted here to determine social sentiment. ShezheaNET filters both public trends and "sub-culture" trends using hashtag growth. The actual content of the trending videos has yet to be manually described at the time of writing this litepaper. In the near future, external AI analytics tool will be able to extract the sentiment of a video, even if it is designed to be ironic or sarcastic. Platform X, on the other hand, can be used as a data point due to its diverse opinion culture and easy-to-read data set, without any manual intervention. Here, public trends are used for the most part, though the name (not the ticker) of a company is also used. In addition to measuring social sentiment, traditional means such as metrics analysis and live data feeds from Bloomberg. With a processing speed of 1.3s from Bloomberg live feed to rank and price the event, the AI is at a much more efficient and higher level than traditional alternatives.