怎么查有做网站的公司,轻量级服务器wordpress,网站网页设计中怎么添加页码信息,寻花问柳-专注做一家男人的网站猪大家好#xff0c;我是带我去滑雪#xff01; 为了给投资者提供更准确的投资建议、帮助政府和监管部门更好地制定相关政策#xff0c;维护市场稳定#xff0c;本文对股民情绪和上证指数之间的关系进行更深入的研究#xff0c;并结合信号分解、优化算法和深度学习对上证指数… 大家好我是带我去滑雪 为了给投资者提供更准确的投资建议、帮助政府和监管部门更好地制定相关政策维护市场稳定本文对股民情绪和上证指数之间的关系进行更深入的研究并结合信号分解、优化算法和深度学习对上证指数进行预测以期更好地理解股市的运行规律为股民提供一定的参考。 选取了2014年12月17日至2024年4月17日的共计2269个交易日(覆盖牛市、熊市及调整期多个市场周期)的上证指数数据建立了基于信号分解和多种深度学习结合的预测模型。在建模过程中针对变分模态分解(VMD)参数难以选择的问题我们利用鲸鱼捕食者算法(MPA)寻优能力强的特点自适应选取VMD的关键参数使VMD分解效果最大化针对预测精度问题我们使用LSTM有效传递和表达经MPA-VMD分解后的上证指数信息使用双向门控循环单元(BiGRU)模型通过正向和反向传播进行信息的交互和整合添加注意力机制(Attention)为特征提供更具针对性的权重分配最大程度上提高预测模型的效率和性能。 最终建立了基于上证指数的VMD-LSTM-BiGRU-Attention预测模型该模型的拟合优度达到了0.98而均方根误差(RMSE)、平均绝对误差(MAE)和平均绝对百分比误差(MAPE)也明显优于本文用于比较的其他模型说明本文模型可为上证指数提供更加精准的预测。不仅如此在相同VMD-LSTM-BiGRU-Attention模型的条件下加入情绪指数与上证指数结合进行预测比单独使用上证指数进行预测的准确率更高。这很好的体现了本文研究股民情绪指数和上证指数之间的关联性是有实际意义的实现了基于股民情绪与上证指数混合模型的股票走向预测且性能优秀可用于上证指数的预测分析对股民及相关研究领域有一定的参考价值和指导作用。 下面开始代码实战
目录
1数据展示
2MPA算法优化的VMD分解
3预测模型 4模型评价指标 1数据展示 选取东方财富网2014年12月17日至2024年4月17日的上证指数数据共计2269个样本将其划分为训练集和测试集其中训练集为数据前80%共计1815个测试集为数据的后20%共计454个。利用python将上证指数数据可视化如图16所示。其中黄色实线为训练集数据粉色虚线为测试集数据。 2MPA算法优化的VMD分解
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
import MPA
from scipy.signal import hilbert
from vmdpy import VMD
f pd.read_csv(E:\工作\硕士\上证指数.csv,encodingANSI)
ff.收盘
# VMD参数
tau 0. # noise-tolerance (no strict fidelity enforcement) # 3 modes
DC 0 # no DC part imposed
init 1 # initialize omegas uniformly
tol 1e-7
#MPA设置参数
num_particles 30 #种群数量
MaxIter 1000 #最大迭代次数
dim 10 #维度
lb -10*np.ones([dim, 1]) #下边界
ub 10*np.ones([dim, 1])#上边界# 计算每个IMF分量的包络熵
import numpy as np
from scipy.signal import hilbert
from scipy.stats import kurtosisdef calculate_entropy(imf):# 计算希尔伯特变换的包络线env np.abs(hilbert(imf))# 归一化包络线env_norm env / np.max(env)# 计算归一化包络线的概率分布p env_norm / np.sum(env_norm)# 计算包络熵entropy -np.sum(p * np.log2(p))# 计算峭度kurt kurtosis(imf)# 返回包络熵和峭度的和return entropy kurt# 定义适应度函数即最大包络熵
def fitness_func(x):if x[1] 0:x[1] np.random.choice([3, 12])u, u_hat, omega VMD(f, int(round(x[0])), tau, int(round(x[1])), DC, init, tol)num_modes u.shape[0]entropy np.zeros(num_modes)for i in range(num_modes):entropy[i] calculate_entropy(u[i,:])# 找到最小的包络熵对应的模态min_entropy_index np.argmin(entropy)min_entropy_mode u[min_entropy_index]# print(最小包络熵对应的模态, min_entropy_index)# x为VMD参数向量# signal为要分解的信号# 分解信号并计算最大包络熵# 返回最大包络熵值return entropy[min_entropy_index]def mpa_optimization(num_particles, dim,fitness_func, MaxIter):# 初始化海洋捕食者位置和速度predators_pos np.zeros((num_particles, num_dimensions))for i in range(num_particles):predators_pos[i, 0] np.random.uniform(500, 3000)predators_pos[i, 1] np.random.uniform(3, 12)predators_vel np.zeros((num_particles, num_dimensions))# 记录每个海洋捕食者的最佳位置和适应度值predators_best_pos np.copy(predators_pos)predators_best_fit np.zeros(num_particles)# 记录整个群体的最佳位置和适应度值global_best_pos np.zeros(num_dimensions)global_best_fit float(inf)# 迭代更新for i in range(max_iter):# 计算每个海洋捕食者的适应度值predators_fitness np.array([fitness_func(p) for p in predators_pos])# 更新每个海洋捕食者的最佳位置和适应度值for j in range(num_particles):if predators_fitness[j] predators_best_fit[j]:predators_best_fit[j] predators_fitness[j]predators_best_pos[j] np.copy(predators_pos[j])# 更新整个群体的最佳位置和适应度值global_best_idx np.argmin(predators_fitness)if predators_fitness[global_best_idx] global_best_fit:global_best_fit predators_fitness[global_best_idx]global_best_pos np.copy(predators_pos[global_best_idx])# 更新每个海洋捕食者的速度和位置for j in range(num_particles):# 计算新速度r1 np.random.rand(num_dimensions)r2 np.random.rand(num_dimensions)cognitive_vel 2.0 * r1 * (predators_best_pos[j] - predators_pos[j])social_vel 2.0 * r2 * (global_best_pos - predators_pos[j])predators_vel[j] predators_vel[j] cognitive_vel social_vel# 更新位置predators_pos[j] predators_pos[j] predators_vel[j]# 记录每次迭代的global_best_pos和global_best_fitglobal_best_pos_list.append(global_best_pos)global_best_fit_list.append(global_best_fit)print(第 str(i) 次迭代)# 返回全局最优# 返回全局最优位置和适应度值return global_best_pos, global_best_fit
# 初始化空列表用于存储每次迭代的global_best_pos和global_best_fit
global_best_pos_list []
global_best_fit_list []
# 使用PSO算法优化VMD参数
num_particles 2
num_dimensions 2 # 假设有2个VMD参数
max_iter 30
best_pos, best_fit mpa_optimization(num_particles, dim,fitness_func, MaxItermax_iter)# 输出结果
print(Best VMD parameters:, best_pos)
print(Best fitness value:, best_fit) 得到的最优参数组合和适应度变化曲线如下 得到的分解序列和中心模态分别为如下所示 在分解结果中IMF1显示了一种相对平滑的趋势可能代表了原始序列中的最低频成分或趋势项。而IMF2显示了略微增加的频率和振幅但依旧比较平滑可能捕捉到了次低频的周期性变化。IMF3-IMF7则展示了随着序号增加频率逐渐增高的成分。随着IMF序号的增加我们可以看到更高频的振荡这表明它们捕获了原始上证指数序列中的更快变化部分。可认为经MPA优化后的VMD在处理非线性和非平稳信号时具有较好的性能它能够提取出不同频率范围的振动模态并具有一定的抗噪能力和平滑性适用于本文上证指数的分解。
3预测模型 根据以上分解过程VMD将原始上证指数序列分解成多个代表不同频率成分的模态。通过对这些模态分别进行预测最终可以实现上证指数的预测。本文之后就将基于VMD分解和深度学习结合的方法对上证指数进行预测。为了选取最适合本文上证指数预测的VMD-深度学习模型在这里我们做了大量实验。首先将上证指数数据集以82的比例划分训练集和测试集用训练集分别训练LSTM单一模型、VMD-LSTM、VMD-LSTM-BiGRU、VMD-LSTM-BiGRU-Attention。碍于篇幅原因这里以VMD-LSTM-BiGRU-Attention模型为例展示各个子模态的预测情况如下所示 通过上面可以发现VMD-LSTM-BiGRU-Attention模型的各个子模态的预测效果均较好预测模型能够较好地跟踪和预测多个频率层面上的序列数据。经训练后分别使用了LSTM单一模型、VMD-LSTM、VMD-LSTM-BiGRU、VMD-LSTM-BiGRU-Attention对测试集进行拟合预测并与实际值进行对比绘制了不同模型预测效果对比图如下图所示 部分代码 from keras import backend as K
from keras.layers import Layerclass Embedding(Layer):def __init__(self, vocab_size, model_dim, **kwargs):self._vocab_size vocab_sizeself._model_dim model_dimsuper(Embedding, self).__init__(**kwargs)def build(self, input_shape):self.embeddings self.add_weight(shape(self._vocab_size, self._model_dim),initializerglorot_uniform,nameembeddings)super(Embedding, self).build(input_shape)def call(self, inputs):if K.dtype(inputs) ! int32:inputs K.cast(inputs, int32)embeddings K.gather(self.embeddings, inputs)embeddings * self._model_dim ** 0.5 # Scalereturn embeddingsdef compute_output_shape(self, input_shape):return input_shape (self._model_dim,)class PositionEncoding(Layer):def __init__(self, model_dim, **kwargs):self._model_dim model_dimsuper(PositionEncoding, self).__init__(**kwargs)def call(self, inputs):seq_length inputs.shape[1]position_encodings np.zeros((seq_length, self._model_dim))for pos in range(seq_length):for i in range(self._model_dim):position_encodings[pos, i] pos / np.power(10000, (i-i%2) / self._model_dim)position_encodings[:, 0::2] np.sin(position_encodings[:, 0::2]) # 2iposition_encodings[:, 1::2] np.cos(position_encodings[:, 1::2]) # 2i1position_encodings K.cast(position_encodings, float32)return position_encodingsdef compute_output_shape(self, input_shape):return input_shape
class Add(Layer):def __init__(self, **kwargs):super(Add, self).__init__(**kwargs)def call(self, inputs):input_a, input_b inputsreturn input_a input_bdef compute_output_shape(self, input_shape):return input_shape[0]class ScaledDotProductAttention(Layer):def __init__(self, maskingTrue, futureFalse, dropout_rate0., **kwargs):self._masking maskingself._future futureself._dropout_rate dropout_rateself._masking_num -2**321super(ScaledDotProductAttention, self).__init__(**kwargs)def mask(self, inputs, masks):masks K.cast(masks, float32)masks K.tile(masks, [K.shape(inputs)[0] // K.shape(masks)[0], 1])masks K.expand_dims(masks, 1)outputs inputs masks * self._masking_numreturn outputsdef future_mask(self, inputs):diag_vals tf.ones_like(inputs[0, :, :])tril tf.linalg.LinearOperatorLowerTriangular(diag_vals).to_dense() future_masks tf.tile(tf.expand_dims(tril, 0), [tf.shape(inputs)[0], 1, 1])paddings tf.ones_like(future_masks) * self._masking_numoutputs tf.where(tf.equal(future_masks, 0), paddings, inputs)return outputsdef call(self, inputs):if self._masking:assert len(inputs) 4, inputs should be set [queries, keys, values, masks].queries, keys, values, masks inputselse:assert len(inputs) 3, inputs should be set [queries, keys, values].queries, keys, values inputsif K.dtype(queries) ! float32: queries K.cast(queries, float32)if K.dtype(keys) ! float32: keys K.cast(keys, float32)if K.dtype(values) ! float32: values K.cast(values, float32)matmul K.batch_dot(queries, tf.transpose(keys, [0, 2, 1])) # MatMulscaled_matmul matmul / int(queries.shape[-1]) ** 0.5 # Scaleif self._masking:scaled_matmul self.mask(scaled_matmul, masks) # Mask(opt.)if self._future:scaled_matmul self.future_mask(scaled_matmul)softmax_out K.softmax(scaled_matmul) # SoftMax# Dropoutout K.dropout(softmax_out, self._dropout_rate)outputs K.batch_dot(out, values)return outputsdef compute_output_shape(self, input_shape):return input_shapeclass MultiHeadAttention(Layer):def __init__(self, n_heads, head_dim, dropout_rate.1, maskingTrue, futureFalse, trainableTrue, **kwargs):self._n_heads n_headsself._head_dim head_dimself._dropout_rate dropout_rateself._masking maskingself._future futureself._trainable trainablesuper(MultiHeadAttention, self).__init__(**kwargs)def build(self, input_shape):self._weights_queries self.add_weight(shape(input_shape[0][-1], self._n_heads * self._head_dim),initializerglorot_uniform,trainableself._trainable,nameweights_queries)self._weights_keys self.add_weight(shape(input_shape[1][-1], self._n_heads * self._head_dim),initializerglorot_uniform,trainableself._trainable,nameweights_keys)self._weights_values self.add_weight(shape(input_shape[2][-1], self._n_heads * self._head_dim),initializerglorot_uniform,trainableself._trainable,nameweights_values)super(MultiHeadAttention, self).build(input_shape)def call(self, inputs):if self._masking:assert len(inputs) 4, inputs should be set [queries, keys, values, masks].queries, keys, values, masks inputselse:assert len(inputs) 3, inputs should be set [queries, keys, values].queries, keys, values inputsqueries_linear K.dot(queries, self._weights_queries) keys_linear K.dot(keys, self._weights_keys)values_linear K.dot(values, self._weights_values)queries_multi_heads tf.concat(tf.split(queries_linear, self._n_heads, axis2), axis0)keys_multi_heads tf.concat(tf.split(keys_linear, self._n_heads, axis2), axis0)values_multi_heads tf.concat(tf.split(values_linear, self._n_heads, axis2), axis0)if self._masking:att_inputs [queries_multi_heads, keys_multi_heads, values_multi_heads, masks]else:att_inputs [queries_multi_heads, keys_multi_heads, values_multi_heads]attention ScaledDotProductAttention(maskingself._masking, futureself._future, dropout_rateself._dropout_rate)att_out attention(att_inputs)outputs tf.concat(tf.split(att_out, self._n_heads, axis0), axis2)return outputsdef compute_output_shape(self, input_shape):return input_shape
def build_model(X_train,modeLSTM,hidden_dim[32,16]):set_my_seed()if modeRNN:#RNNmodel Sequential()model.add(SimpleRNN(hidden_dim[0],return_sequencesTrue, input_shape(X_train.shape[-2],X_train.shape[-1])))model.add(SimpleRNN(hidden_dim[1])) model.add(Dense(1))elif modeMLP:model Sequential()model.add(Dense(hidden_dim[0],activationrelu,input_shape(X_train.shape[-1],)))model.add(Dense(hidden_dim[1],activationrelu))model.add(Dense(1))elif modeLSTM:# LSTMmodel Sequential()model.add(LSTM(hidden_dim[0],return_sequencesTrue, input_shape(X_train.shape[-2],X_train.shape[-1])))model.add(LSTM(hidden_dim[1]))model.add(Dense(1))elif modeGRU:#GRUmodel Sequential()model.add(GRU(hidden_dim[0],return_sequencesTrue, input_shape(X_train.shape[-2],X_train.shape[-1])))model.add(GRU(hidden_dim[1]))model.add(Dense(1))elif modeCNN:#一维卷积model Sequential()model.add(Conv1D(hidden_dim[0],17,activationrelu,input_shape(X_train.shape[-2],X_train.shape[-1])))model.add(GlobalAveragePooling1D())model.add(Flatten())model.add(Dense(hidden_dim[1],activationrelu))model.add(Dense(1))elif modeCNNLSTM: model Sequential()model.add(Conv1D(filtershidden_dim[0], kernel_size3, paddingsame,activationrelu))model.add(MaxPooling1D(pool_size2))model.add(LSTM(hidden_dim[1]))model.add(Dense(1))elif modeBiLSTM:model Sequential()model.add(Bidirectional(LSTM(hidden_dim[0],return_sequencesTrue, input_shape(X_train.shape[-2],X_train.shape[-1]))))model.add(Bidirectional(LSTM(hidden_dim[1])))model.add(Dense(1))elif modeBiGRU:model Sequential()model.add(Bidirectional(GRU(hidden_dim[0],return_sequencesTrue, input_shape(X_train.shape[-2],X_train.shape[-1]))))model.add(Bidirectional(GRU(hidden_dim[1])))model.add(Dense(1))elif modeBiLSTMAttention:inputs Input(nameinputs,shape[X_train.shape[-2],X_train.shape[-1]], dtypefloat64)attention_probs Dense(32, activationsoftmax, nameattention_vec)(inputs)attention_mul Multiply()([inputs, attention_probs])mlp Dense(64)(attention_mul) #原始的全连接gruBidirectional(LSTM(32))(mlp)mlp Dense(16,activationrelu)(gru)output Dense(1)(mlp)model Model(inputs[inputs], outputsoutput)elif modeAttention:inputs Input(nameinputs,shape[X_train.shape[-2],X_train.shape[-1]], dtypefloat32)attention_probs Dense(hidden_dim[0], activationsoftmax, nameattention_vec)(inputs)attention_mul Multiply()([inputs, attention_probs])mlp Dense(hidden_dim[1])(attention_mul) #原始的全连接flaFlatten()(mlp)output Dense(1)(fla)model Model(inputs[inputs], outputsoutput) elif modeBiGRUAttention:inputs Input(nameinputs,shape[X_train.shape[-2],X_train.shape[-1]], dtypefloat64)attention_probs Dense(32, activationsoftmax, nameattention_vec)(inputs)attention_mul Multiply()([inputs, attention_probs])mlp Dense(64)(attention_mul) #原始的全连接gruBidirectional(GRU(32))(mlp)mlp Dense(16,activationrelu)(gru)output Dense(1)(mlp)model Model(inputs[inputs], outputsoutput)elif modeMultiHeadAttention: inputs Input(shape[X_train.shape[-2],X_train.shape[-1]], nameinputs)#masks Input(shape(X_train.shape[-2],), namemasks)encodings PositionEncoding(X_train.shape[-2])(inputs)encodings Add()([inputs, encodings])x MultiHeadAttention(8, hidden_dim[0],maskingFalse)([encodings, encodings, encodings])x GlobalAveragePooling1D()(x)x Dropout(0.2)(x)x Dense(hidden_dim[1], activationrelu)(x)outputs Dense(1)(x)model Model(inputs[inputs], outputsoutputs)elif modeBiGRUMAttention: inputs Input(shape[X_train.shape[-2],X_train.shape[-1]], nameinputs)encodings PositionEncoding(X_train.shape[-2])(inputs)encodings Add()([inputs, encodings])x MultiHeadAttention(8, hidden_dim[0],maskingFalse)([encodings, encodings, encodings])
# x GlobalAveragePooling1D()(x)x Bidirectional(GRU(32))(x)x Dropout(0.2)(x)output Dense(1)(x)model Model(inputs[inputs], outputsoutput)# elif modeBiGRUAttention:
# inputs Input(nameinputs,shape[max_words,], dtypefloat64)
# x Embedding(top_words, input_lengthmax_words, output_dimembed_dim)(inputs)
# x Bidirectional(GRU(32,return_sequencesTrue))(x)
# x MultiHeadAttention(2, key_dimembed_dim)(x,x,x)
# x Bidirectional(GRU(32))(x)
# x Dropout(0.2)(x)
# output Dense(num_labels, activationsoftmax)(x)
# model Model(inputs[inputs], outputsoutput)model.compile(optimizerAdam, lossmse,metrics[tf.keras.metrics.RootMeanSquaredError(),mape,mae])return modeldef plot_loss(hist,imfname):plt.subplots(1,4,figsize(16,2), dpi600)for i,key in enumerate(hist.history.keys()):nint(str(14)str(i1))plt.subplot(n)plt.plot(hist.history[key], k, labelfTraining {key})plt.title(f{imfname} Training {key})plt.xlabel(Epochs)plt.ylabel(key)plt.legend()plt.tight_layout()plt.show() 4模型评价指标 为了直观展示结果绘制评价指标柱状图。如下图所示 需要数据集的家人们可以去百度网盘永久有效获取
链接https://pan.baidu.com/s/16Pp57kAbC3xAqPylyfQziA?pwd2138 提取码2138 更多优质内容持续发布中请移步主页查看。 点赞关注,下次不迷路
