Hasashen Farashin Kudi na Dijital Ta Amfani da Injin Koyo

Teburin Abubuwan Ciki

1. Gabatarwa

Kasuwar kudi na dijital ta sami girma wanda ba a taɓa ganin irinsa ba tun daga 2017, tare da girman kasuwa ya kai sama da dala biliyan 800 a cikin Janairu 2018. Wannan binciken yana magance hasashen rashin ingancin kasuwa ta hanyar amfani da algorithms na injin koyo na zamani don hasashen farashin kudi na dijital da kuma samar da riba mai yawa ta hanyar dabarun ciniki na algorithm.

2. Hanyar Aiki

2.1 Tattara Bayanai

Binciken ya bincika bayanan yau da kullun na kudi na dijital 1,681 daga Nuwamba 2015 zuwa Afrilu 2018. Bayanan sun haɗa da motsin farashi, yawan ciniki, da ma'aunin girman kasuwa a cikin musayar da yawa ciki har da Binance, Upbit, da Kraken.

2.2 Samfuran Injin Koyo

An kimanta manyan samfura guda uku:

• Aiwar haɓakar yanke shawara na gradient guda biyu (XGBoost, LightGBM)
• Cibiyoyin sadarwa na juyayi na Dogon Lokaci Gajere (LSTM)

2.3 Aiwar da Dabarun Ciniki

An gina fayafayan saka hannun jari bisa ga hasashen samfurin, tare da auna aikin ta hanyar dawowar saka hannun jari (ROI) idan aka kwatanta da ma'auni na yau da kullun ciki har da dabarun saye da riƙe.

3. Aiwar da Fasaha

3.1 Tsarin Lissafi

Matsalar hasashen farashi za a iya tsara ta azaman aikin hasashen jerin lokaci. Bari $P_t$ ya wakilci farashi a lokacin $t$, kuma $X_t$ ya wakilci sifofin fasali ciki har da farashin tarihi, girma, da alamomin fasaha. Samfurin hasashe yana neman koyo:

$P_{t+1} = f(X_t, X_{t-1}, ..., X_{t-n}) + \epsilon_t$

inda $f$ ke wakiltar samfurin injin koyo kuma $\epsilon_t$ shine kalmar kuskure.

3.2 Cikakkun Bayanai na Algorithm

Haɓakar gradient yana gina tarin samfuran hasashe masu rauni, yawanci bishiyoyin yanke shawara, a cikin yanayin mataki. Algorithm yana rage aikin asara $L$ ta hanyar ƙara bishiyoyin da ke hasashen ragowar bishiyoyin da suka gabata:

$F_m(x) = F_{m-1}(x) + \gamma_m h_m(x)$

inda $h_m(x)$ shine mai koyo na tushe kuma $\gamma_m$ shine girman mataki.

4. Sakamakon Gwaji

Binciken ya nuna cewa dabarun ciniki masu taimakon injin koyo sun fi ma'auni na yau da kullun akai-akai. Manyan binciken sun haɗa da:

• Duk samfura guda uku sun samar da dawowar da ba ta da kyau
• Algorithms na haɓakar gradient sun nuna mafi girman aiki a yawancin yanayi
• Cibiyoyin sadarwa na LSTM sun kama abubuwan dogaro na ɗan lokaci masu sarƙaƙƙiya amma suna buƙatar ƙarin albarkatun lissafi
• Hanyoyin algorithm masu sauƙi sun yi hasashen juyin halittar kasuwa na ɗan gajeren lokaci yadda ya kamata

5. Aiwar da Lambar

A ƙasa akwai sauƙaƙƙen aiwar da hanyar haɓakar gradient a cikin Python:

import xgboost as xgb
import pandas as pd
from sklearn.metrics import mean_squared_error

# Aikin injiniyan fasali
def create_features(df):
    df['price_lag1'] = df['price'].shift(1)
    df['volume_lag1'] = df['volume'].shift(1)
    df['price_rolling_mean'] = df['price'].rolling(window=7).mean()
    return df.dropna()

# Horar da samfuri da hasashe
model = xgb.XGBRegressor(
    n_estimators=100,
    max_depth=6,
    learning_rate=0.1
)

# Ana ɗauka X_train, y_train an shirya fasali da hari
model.fit(X_train, y_train)
predictions = model.predict(X_test)

6. Ayyuka na Gaba

Nasarar injin koyo a cikin hasashen kudi na dijital ya buɗe hanyoyi da yawa na gaba:

• Haɗa madogaran bayanai madadin (ra'ayin jama'a a kafofin sada zumunta, ma'aunin blockchain)
• Haɓakar samfuran gauraye waɗanda suka haɗa bincike na asali da na fasaha
• Aiwatar da gine-ginen canzawa don ingantaccen samfurin jeri
• Tsarin ciniki na ainihin lokaci tare da tsare-tsaren sarrafa haɗari
• Ingantaccen fayafayan saka hannun jari na kayan dukiya wanda ya haɗa da kayan al'ada da na crypto

7. Nassoshi

1. ElBahrawy, A., et al. (2017). Evolutionary dynamics of the cryptocurrency market. Royal Society Open Science.
2. Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. KDD '16.
3. Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation.
4. Brown, T. B., et al. (2020). Language Models are Few-Shot Learners. NeurIPS.
5. Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance.