Анализ важнеших СТАТИСТИЧЕСКИХ характеристик паттерна и выбор метода торговли по нему. - страница 7
Вы упускаете торговые возможности:
- Бесплатные приложения для трейдинга
- 8 000+ сигналов для копирования
- Экономические новости для анализа финансовых рынков
Регистрация
Вход
Вы принимаете политику сайта и условия использования
Если у вас нет учетной записи, зарегистрируйтесь
Форум по трейдингу, автоматическим торговым системам и тестированию торговых стратегий
FOREX - Тенденции, прогнозы и следствия 2015
-Aleks-, 2015.06.21 12:29
Look in the code. The method is quite simple. Set the length of the current pattern, find similar patterns in history (for example, use correlation as the distance between patterns), predict the behavior of the price in the future from past patterns. This is essentially the same clustering, or RBF, or SVM, or GRNN. It all depends on how we measure the distance from the current pattern. Read on GRNN and Bayes. There, the theory of predictions is described in terms of statistical distributions . Written about GRNN and the above methods of predictions a lot, but it comes down to one simple formula:
the prediction is y = SUM y [k] * exp (-d [k] / 2s ^ 2) / SUM exp (-d [k] / 2s ^ 2)
where y [k] is the k-th past pattern, d [k] is the distance from the k-th pattern to the current one. If the distances have a Gaussian distribution, then d [k] = (x - x [k]) ^ 2. For an arbitrary (super Gaussian) distribution, d [k] = | x - x [k] | ^ p, where you choose p, depending on whether you want to give more weight to the nearest neighbors (large p), or to give almost all neighbors the same weight (small p) as under socialism. For p = 0, we have complete socialism.
After getting acquainted with the nearest neighbors and GRNN, the following obvious question will arise. And how to measure the distance between the current pattern and past patterns. If you take into account the distortions along the time axis. This is where the dog is buried.
perhaps this will help https://en.wikipedia.org/wiki/Hellinger_distance
https://dl.acm.org/citation.cfm?id=1288866