股票市场的不确定性和复杂性使得股票预测成为一项具有挑战性的任务。鉴于金融文本在股票预测中的潜在价值,采用词典法和BERT双向长短期记忆模型(bidirectional encoder representations from transformers-bidirectional long short-te...股票市场的不确定性和复杂性使得股票预测成为一项具有挑战性的任务。鉴于金融文本在股票预测中的潜在价值,采用词典法和BERT双向长短期记忆模型(bidirectional encoder representations from transformers-bidirectional long short-term memory,BERT-BiLSTM)对在线财经新闻提取情感特征,构建了融合情感特征和股票交易特征的股指预测模型。实验对比了融合情感特征前后模型的预测能力,并探讨了不同模型、不同时间周期下预测能力的差异。实验结果表明,融合词典法和深度学习技术提取的情感特征均能提升各模型股指预测的准确率。LSTM模型相较其他实验模型在融合情感特征前后的股指预测上均表现较好。进一步的时间跨度分析表明,股指预测模型在较短的时间跨度上对股票指数涨跌的预测能力更强。为验证股指预测模型的实际价值,对沪深300指数的牛熊市和震荡市进行回测分析,结合LSTM模型和深度Q网络(deep Q-network,DQN)原理,对比了传统均线策略以及结合DQN强化学习算法后股指回测差异。回测结果表明,相比于单一的传统交易策略,结合传统交易策略和深度学习方法的股票指数预测模型在牛熊市及震荡市中均保证了正的夏普比例和累积收益率,并有效控制了最大回撤,显示出更强的市场适应性和盈利能力。展开更多
Similar to the method of continuum mechanics, the variation of the price of index futures is viewed to be continuous and regular. According to the characteristic of index futures, a basic equation of price of index fu...Similar to the method of continuum mechanics, the variation of the price of index futures is viewed to be continuous and regular. According to the characteristic of index futures, a basic equation of price of index futures was established. It is a differential equation, its solution shows that the relation between time and price forms a logarithmic circle. If the time is thought of as the probability of its corresponding price, then such a relation is perfectly coincided with the main assumption of the famous formula of option pricing, based on statistical theory, established by Black and Scholes, winner of 1997 Nobel’ prize on economy. In that formula, the probability of price of basic assets (they stand for index futures here) is assummed to be a logarithmic normal distribution. This agreement shows that the same result may be obtained by two analytic methods with different bases. However, the result, given by assumption by Black_Scholes, is derived from the solution of the differential equation.展开更多
文摘股票市场的不确定性和复杂性使得股票预测成为一项具有挑战性的任务。鉴于金融文本在股票预测中的潜在价值,采用词典法和BERT双向长短期记忆模型(bidirectional encoder representations from transformers-bidirectional long short-term memory,BERT-BiLSTM)对在线财经新闻提取情感特征,构建了融合情感特征和股票交易特征的股指预测模型。实验对比了融合情感特征前后模型的预测能力,并探讨了不同模型、不同时间周期下预测能力的差异。实验结果表明,融合词典法和深度学习技术提取的情感特征均能提升各模型股指预测的准确率。LSTM模型相较其他实验模型在融合情感特征前后的股指预测上均表现较好。进一步的时间跨度分析表明,股指预测模型在较短的时间跨度上对股票指数涨跌的预测能力更强。为验证股指预测模型的实际价值,对沪深300指数的牛熊市和震荡市进行回测分析,结合LSTM模型和深度Q网络(deep Q-network,DQN)原理,对比了传统均线策略以及结合DQN强化学习算法后股指回测差异。回测结果表明,相比于单一的传统交易策略,结合传统交易策略和深度学习方法的股票指数预测模型在牛熊市及震荡市中均保证了正的夏普比例和累积收益率,并有效控制了最大回撤,显示出更强的市场适应性和盈利能力。
文摘Similar to the method of continuum mechanics, the variation of the price of index futures is viewed to be continuous and regular. According to the characteristic of index futures, a basic equation of price of index futures was established. It is a differential equation, its solution shows that the relation between time and price forms a logarithmic circle. If the time is thought of as the probability of its corresponding price, then such a relation is perfectly coincided with the main assumption of the famous formula of option pricing, based on statistical theory, established by Black and Scholes, winner of 1997 Nobel’ prize on economy. In that formula, the probability of price of basic assets (they stand for index futures here) is assummed to be a logarithmic normal distribution. This agreement shows that the same result may be obtained by two analytic methods with different bases. However, the result, given by assumption by Black_Scholes, is derived from the solution of the differential equation.
