1、摘要
本文主要讲解：麻雀算法SSA优化LSTM长短期记忆网络实现分类算法
主要思路：
准备一份分类数据，数据介绍在第二章准备好麻雀算法SSA，要用随机数据跑起来用lstm把分类数据跑起来将lstm的超参数交给SSA去优化优化完的最优参数给lstm去做最后一次训练
2、数据介绍
Cll：出料量
Lsp：量水平
Djzsp：电解质水平
Djwd：工作温度
Fzb：分子比
Fe：铁含量
Si:硅含量
Ludiyajiang：压降
Ddlsp：打点量水平
Avv：平均电压
wv：工作电压
avaev：平均故障发生时的电压
ae：故障发生标签（0：未发生，1：发生）
除分子比、压降、打点量水平外其余数据均是一天一采集，数据异常时发生故障，故障电压高于工作电压。
数据下载链接
3、相关技术
麻雀搜索算法(Sparrow Search Algorithm, SSA)是一种新型的群智能优化算法，在提出，主要是受麻雀的觅食行为和反捕食行为的启发 ，以下是一些图片，可加深你的理解：
4、完整代码和步骤
此代码的依赖环境如下：
tensorflow==2.5.0numpy==1.19.5keras==2.6.0matplotlib==3.5.2
麻雀算法用随机数据跑起来的代码：
# -*- coding: utf-8 -*-import numpy as npimport randomimport matplotlib.pyplot as pltdef fun(x):a = np.sum(x ** 2)return adef Bounds(x, lb, ub):temp = x.reshape(-1, 2)I = temp < lbtemp[I] = lb[I]J = temp > ubtemp[J] = ub[J]return tempdef SSA(M, pop, dim, P_percent, c, d):# M 迭代次数 SSA(1000, 20, 2, 0.2, -10, 10)# pop 麻雀种群数量# dim 寻优维度# P_percent 麻雀在生产者的比例# c d分别是寻优范围的最小值与最大值pNum = round(pop * P_percent) # pNum是生产者lb = c * np.ones((1, dim)) # 下边界ub = d * np.ones((1, dim)) # 上边界x = np.zeros((pop, dim))fit = np.zeros((pop, 1))# 种群初始化for i in range(pop):x[i, :] = lb + (ub - lb) * np.random.rand(1, dim)fit[i] = fun(x[i, :])pFit = fit.copy()pX = x.copy()fMin = np.min(fit)bestI = np.argmin(fit)bestX = x[bestI, :].copy()Convergence_curve = np.zeros((M,))trace = np.zeros((M, dim))for t in range(M):sortIndex = np.argsort(pFit.reshape(-1, )).reshape(-1, )fmax = np.max(pFit)B = np.argmax(pFit)worse = x[B, :].copy()r2 = np.random.rand()## 这一部分为发现者（探索者）的位置更新if r2 < 0.8: # %预警值较小，说明没有捕食者出现for i in range(pNum): # r2小于0.8时发现者改变位置r1 = np.random.rand()x[sortIndex[i], :] = pX[sortIndex[i], :] * np.exp(-i / (r1 * M))x[sortIndex[i], :] = Bounds(x[sortIndex[i], :], lb, ub)temp = fun(x[sortIndex[i], :])fit[sortIndex[i]] = temp # 计算新的适应度值else: # 预警值较大，说明有捕食者出现威胁到了种群的安全，需要去其它地方觅食for i in range(pNum): # r2大于0.8时发现者改变位置r1 = np.random.rand()x[sortIndex[i], :] = pX[sortIndex[i], :] + np.random.normal() * np.ones((1, dim))x[sortIndex[i], :] = Bounds(x[sortIndex[i], :], lb, ub)fit[sortIndex[i]] = fun(x[sortIndex[i], :]) # 计算新的适应度值bestII = np.argmin(fit)bestXX = x[bestII, :].copy()##这一部分为加入者（追随者）的位置更新for i in range(pNum + 1, pop): # 剩下的个体变化A = np.floor(np.random.rand(1, dim) * 2) * 2 - 1if i > pop / 2: # 这个代表这部分麻雀处于十分饥饿的状态（因为它们的能量很低，也是是适应度值很差），需要到其它地方觅食x[sortIndex[i], :] = np.random.normal() * np.exp((worse - pX[sortIndex[i], :]) / (i ** 2))else: # 这一部分追随者是围绕最好的发现者周围进行觅食，其间也有可能发生食物的争夺，使其自己变成生产者x[sortIndex[i], :] = bestXX + np.abs(pX[sortIndex[i], :] - bestXX).dot(A.T * (A * A.T) ** (-1)) * np.ones((1, dim))x[sortIndex[i], :] = Bounds(x[sortIndex[i], :], lb, ub) # 判断边界是否超出fit[sortIndex[i]] = fun(x[sortIndex[i], :]) # 计算适应度值# 这一部分为意识到危险（注意这里只是意识到了危险，不代表出现了真正的捕食者）的麻雀的位置更新c = random.sample(range(sortIndex.shape[0]),sortIndex.shape[0]) # 这个的作用是在种群中随机产生其位置（也就是这部分的麻雀位置一开始是随机的，意识到危险了要进行位置移动，b = sortIndex[np.array(c)[0:round(pop * 0.2)]].reshape(-1, )for j in range(b.shape[0]):if pFit[sortIndex[b[j]]] > fMin: # 处于种群外围的麻雀的位置改变x[sortIndex[b[j]], :] = bestX + np.random.normal(1, dim) * (np.abs(pX[sortIndex[b[j]], :] - bestX))else: # 处于种群中心的麻雀的位置改变x[sortIndex[b[j]], :] = pX[sortIndex[b[j]], :] + (2 * np.random.rand() - 1) * (np.abs(pX[sortIndex[b[j]], :] - worse)) / (pFit[sortIndex[b[j]]] - fmax + 1e-50)x[sortIndex[b[j]], :] = Bounds(x[sortIndex[b[j]], :], lb, ub)fit[sortIndex[b[j]]] = fun(x[sortIndex[b[j]], :]) # 计算适应度值# 这部分是最终的最优解更新for i in range(pop):if fit[i] < pFit[i]:pFit[i] = fit[i].copy()pX[i, :] = x[i, :].copy()if pFit[i] < fMin:fMin = pFit[i, 0].copy()bestX = pX[i, :].copy()trace[t, :] = bestXConvergence_curve[t] = fMinreturn bestX, fMin, Convergence_curve, trace# In[]bestX, fMin, Convergence_curve, trace = SSA(1000, 20, 2, 0.2, -10, 10)plt.figure()plt.plot(Convergence_curve)plt.show()
代码输出如下：
lstm分类算法实现：
import osimport matplotlib.pyplot as pltimport pandas as pdfrom tensorflow.python.keras import Sequentialfrom tensorflow.python.keras.layers import Dense, CuDNNLSTMfrom tensorflow.python.keras.layers import Dropoutfrom tensorflow.python.keras.models import Sequentialfrom numpy.random import seedseed(7)os.chdir(r'C:\Projects\old\判断异常是否发生')train_path = 'data.csv'# train_path = '15.xlsx'usecols = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]df = pd.read_csv(train_path, usecols=usecols)df['ae'] = df['ae'].map(lambda x: 1 if x >= 1 else 0)df.fillna(0, inplace=True)train_size = int(len(df) * 0.9)train = df.iloc[:train_size, :]test = df.iloc[train_size:, :]X_train = train.loc[:, train.columns != 'ae'].values # converts the df to a numpy arrayy_train = train['ae'].valuesX_train = X_train.astype(float)X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1))print(X_train.shape, y_train.shape)X_test = test.loc[:, test.columns != 'ae'].values # converts the df to a numpy arrayy_test = test['ae'].valuesX_test = X_test.reshape((X_test.shape[0], X_test.shape[1], 1))X_test = X_test.astype(float)print(X_test.shape, X_test.shape)def create_model(input_length):model = Sequential()model.add(CuDNNLSTM(units=50, return_sequences=True, input_shape=(input_length, 1)))model.add(Dropout(0.2))model.add(CuDNNLSTM(units=50, return_sequences=False))model.add(Dropout(0.2))model.add(Dense(1, activation='sigmoid'))pile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])model.summary()return modelmodel = create_model(len(X_train[0]))hist = model.fit(X_train, y_train, batch_size=64, validation_split=0.2, epochs=2, shuffle=False, verbose=1)plt.plot(hist.history['accuracy'], label='acc')plt.plot(hist.history['val_accuracy'], label='val_acc')plt.legend()plt.show()plt.plot(hist.history['loss'], label='loss')plt.legend()plt.show()
分类算法效果如下：
损失图如下：
合并麻雀算法和lstm算法，用麻雀算法SSA优化LSTM长短期记忆网络实现分类算法
from random import seedimport matplotlibimport matplotlib.pyplot as pltimport numpy as npimport pandas as pdfrom sklearn.metrics import precision_score, recall_score, classification_reportfrom tensorflow.python.framework.random_seed import set_random_seedfrom tensorflow.python.keras.layers import CuDNNLSTMfrom tensorflow.python.keras.layers import Dense, Dropoutfrom tensorflow.python.keras.models import Sequential# 这两行代码解决 plt 中文显示的问题plt.rcParams['font.sans-serif'] = ['SimHei']plt.rcParams['axes.unicode_minus'] = Falsematplotlib.rcParams['font.sans-serif'] = ['SimHei'] # 用黑体显示中文set_random_seed(11)seed(7)def pdReadCsv(file, sep):try:data = pd.read_csv(file, sep=sep, encoding='utf-8', error_bad_lines=False, engine='python')return dataexcept:data = pd.read_csv(file, sep=sep, encoding='gbk', error_bad_lines=False, engine='python')return dataclass SSA():def __init__(self, func, n_dim=None, pop_size=20, max_iter=50, lb=-512, ub=512, verbose=False):self.func = funcself.n_dim = n_dim # dimension of particles, which is the number of variables of funcself.pop = pop_size # number of particlesP_percent = 0.2 # # 生产者的人口规模占总人口规模的20%D_percent = 0.1 # 预警者的人口规模占总人口规模的10%self.pNum = round(self.pop * P_percent) # 生产者的人口规模占总人口规模的20%self.warn = round(self.pop * D_percent) # 预警者的人口规模占总人口规模的10%self.max_iter = max_iter # max iterself.verbose = verbose # print the result of each iter or notself.lb, self.ub = np.array(lb) * np.ones(self.n_dim), np.array(ub) * np.ones(self.n_dim)assert self.n_dim == len(self.lb) == len(self.ub), 'dim == len(lb) == len(ub) is not True'assert np.all(self.ub > self.lb), 'upper-bound must be greater than lower-bound'self.X = np.random.uniform(low=self.lb, high=self.ub, size=(self.pop, self.n_dim))self.Y = [self.func(self.X[i]) for i in range(len(self.X))] # y = f(x) for all particlesself.pbest_x = self.X.copy() # personal best location of every particle in historyself.pbest_y = [np.inf for i in range(self.pop)] # best image of every particle in historyself.gbest_x = self.pbest_x.mean(axis=0).reshape(1, -1) # global best location for all particlesself.gbest_y = np.inf # global best y for all particlesself.gbest_y_hist = [] # gbest_y of every iterationself.update_pbest()self.update_gbest()## record verbose valuesself.record_mode = Falseself.record_value = {'X': [], 'V': [], 'Y': []}self.best_x, self.best_y = self.gbest_x, self.gbest_y # history reasons, will be deprecatedself.idx_max = 0self.x_max = self.X[self.idx_max, :]self.y_max = self.Y[self.idx_max]def cal_y(self, start, end):# calculate y for every x in Xfor i in range(start, end):self.Y[i] = self.func(self.X[i])# return self.Ydef update_pbest(self):'''personal best'''for i in range(len(self.Y)):if self.pbest_y[i] > self.Y[i]:self.pbest_x[i] = self.X[i]self.pbest_y[i] = self.Y[i]def update_gbest(self):idx_min = self.pbest_y.index(min(self.pbest_y))if self.gbest_y > self.pbest_y[idx_min]:self.gbest_x = self.X[idx_min, :].copy()self.gbest_y = self.pbest_y[idx_min]def find_worst(self):self.idx_max = self.Y.index(max(self.Y))self.x_max = self.X[self.idx_max, :]self.y_max = self.Y[self.idx_max]def update_finder(self):self.idx = sorted(enumerate(self.Y), key=lambda x: x[1])self.idx = [self.idx[i][0] for i in range(len(self.idx))]# 这一部位为发现者（探索者）的位置更新if r2 < 0.8: # 预警值较小，说明没有捕食者出现for i in range(self.pNum):r1 = np.random.rand(1)self.X[self.idx[i], :] = self.X[self.idx[i], :] * np.exp(-(i) / (r1 * self.max_iter)) # 对自变量做一个随机变换self.X = np.clip(self.X, self.lb, self.ub) # 对超过边界的变量进行去除# X[idx[i], :] = Bounds(X[idx[i], :], lb, ub) # 对超过边界的变量进行去除# fit[sortIndex[0, i], 0] = func(X[sortIndex[0, i], :]) # 算新的适应度值elif r2 >= 0.8: # 预警值较大，说明有捕食者出现威胁到了种群的安全，需要去其它地方觅食for i in range(self.pNum):Q = np.random.rand(1) # 也可以替换成 np.random.normal(loc=0, scale=1.0, size=1)self.X[self.idx[i], :] = self.X[self.idx[i], :] + Q * np.ones((1, self.n_dim)) # Q是服从正态分布的随机数。L表示一个1×d的矩阵self.cal_y(0, self.pNum)def update_follower(self):# 这一部位为加入者（追随者）的位置更新for ii in range(self.pop - self.pNum):i = ii + self.pNumA = np.floor(np.random.rand(1, self.n_dim) * 2) * 2 - 1best_idx = self.Y[0:self.pNum].index(min(self.Y[0:self.pNum]))bestXX = self.X[best_idx, :]if i > self.pop / 2:Q = np.random.rand(1)self.X[self.idx[i], :] = Q * np.exp((self.x_max - self.X[self.idx[i], :]) / np.square(i))else:self.X[self.idx[i], :] = bestXX + np.dot(np.abs(self.X[self.idx[i], :] - bestXX),1 / (A.T * np.dot(A, A.T))) * np.ones((1, self.n_dim))self.X = np.clip(self.X, self.lb, self.ub) # 对超过边界的变量进行去除# X[self.idx[i],:] = Bounds(X[self.idx[i],lb,ub)# fit[self.idx[i],0] = func(X[self.idx[i], :])self.cal_y(self.pNum, self.pop)def detect(self):arrc = np.arange(self.pop)c = np.random.permutation(arrc) # 随机排列序列b = [self.idx[i] for i in c[0: self.warn]]e = 10e-10for j in range(len(b)):if self.Y[b[j]] > self.gbest_y:self.X[b[j], :] = self.gbest_y + np.random.rand(1, self.n_dim) * np.abs(self.X[b[j], :] - self.gbest_y)self.X = np.clip(self.X, self.lb, self.ub) # 对超过边界的变量进行去除self.Y[b[j]] = self.func(self.X[b[j]])def run(self, max_iter=None):self.max_iter = max_iter or self.max_iterfor iter_num in range(self.max_iter):self.update_finder() # 更新发现者位置self.find_worst() # 取出最大的适应度值和最差适应度的Xself.update_follower() # 更新跟随着位置self.update_pbest()self.update_gbest()self.detect()self.update_pbest()self.update_gbest()self.gbest_y_hist.append(self.gbest_y)return self.best_x, self.best_yimport osos.chdir(r'C:\Projects\old\判断异常是否发生')train_path = 'data.csv'# train_path = '15.xlsx'usecols = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]df = pd.read_csv(train_path, usecols=usecols)df['ae'] = df['ae'].map(lambda x: 1 if x >= 1 else 0)df.fillna(0, inplace=True)train_size = int(len(df) * 0.9)train = df.iloc[:train_size, :]test = df.iloc[train_size:, :]X_train = train.loc[:, train.columns != 'ae'].values # converts the df to a numpy arrayy_train = train['ae'].valuesX_train = X_train.astype(float)X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1))print(X_train.shape, y_train.shape)X_test = test.loc[:, test.columns != 'ae'].values # converts the df to a numpy arrayy_test = test['ae'].valuesX_test = X_test.astype(float)X_test = X_test.reshape((X_test.shape[0], X_test.shape[1], 1))print(X_test.shape, X_test.shape)precisions = []recalls = []accuracys = []def create_model(units, dropout):model = Sequential()model.add(CuDNNLSTM(units=units, return_sequences=True, input_shape=(len(X_train[0]), 1)))model.add(Dropout(dropout))model.add(CuDNNLSTM(units=units, return_sequences=False))model.add(Dropout(dropout))model.add(Dense(1, activation='sigmoid'))pile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])return modeldef f13(x):epochs = int(x[0])units = int(x[1])dropout = x[2]batch_size = int(x[3])model = create_model(units, dropout)model.fit(X_train, y_train, batch_size=batch_size, validation_split=0.2, epochs=epochs, shuffle=False, verbose=1)y_pred = model.predict(X_test)y_pred_int = np.argmax(y_pred, axis=1)precision = precision_score(y_test, y_pred_int, average='macro')recall = recall_score(y_test, y_pred_int, average='macro')print('recall')print(recall)recalls.append(recall)print('precision')print(precision)t = classification_report(y_test, y_pred_int, target_names=['0', '1'], output_dict=True)accuracy = t['accuracy']print('accuracy')print(accuracy)accuracys.append(accuracy)precisions.append(precision)return 1 - precision'''根据WIFI数据和成绩数据使用麻雀算法加lgb做一个挂科预测pr曲线，roc曲线和auc值'''if __name__ == '__main__':# 用于优化的四个参数范围 epochs units dropout batch_sizeup_params = [8, 55, 0.55, 55]low_params = [1, 5, 0.05, 5]# 开始优化ssa = SSA(f13, n_dim=4, pop_size=10, max_iter=2, lb=low_params, ub=up_params)ssa.run()print('best_params is ', ssa.gbest_x)print('best_precision is', 1 - ssa.gbest_y)print('best_accuracy is', max(accuracys))print('best_recall is', max(recalls))epochs = int(ssa.gbest_x[0])units = int(ssa.gbest_x[1])dropout = ssa.gbest_x[2]batch_size = int(ssa.gbest_x[3])model = create_model(units, dropout)model.fit(X_train, y_train, batch_size=batch_size, validation_split=0.2, epochs=epochs, shuffle=False, verbose=1)y_pred = model.predict(X_test)y_pred_int = np.argmax(y_pred, axis=1)precision = precision_score(y_test, y_pred_int, average='macro')recall = recall_score(y_test, y_pred_int, average='macro')print('recall')print(recall)recalls.append(recall)print('precision')print(precision)t = classification_report(y_test, y_pred_int, target_names=['0', '1'], output_dict=True)accuracy = t['accuracy']print('accuracy')print(accuracy)
输出如下
D:\Program\CONDA\python.exe D:/Program/JacksonProject/SSA/SSA_LSTM_CLASS.py-03-06 22:51:51.834759: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudart64_110.dll-03-06 22:51:53.071157: I tensorflow/compiler/jit/:41] Not creating XLA devices, tf_xla_enable_xla_devices not set-03-06 22:51:53.071890: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library nvcuda.dll(880, 12, 1) (880,)(98, 12, 1) (98, 12, 1)-03-06 22:51:53.091260: I tensorflow/core/common_runtime/gpu/:1720] Found device 0 with properties: pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3070 computeCapability: 8.6coreClock: 1.755GHz coreCount: 46 deviceMemorySize: 8.00GiB deviceMemoryBandwidth: 417.29GiB/s-03-06 22:51:53.091462: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudart64_110.dll-03-06 22:51:53.097473: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublas64_11.dll-03-06 22:51:53.097585: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublasLt64_11.dll-03-06 22:51:53.100620: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cufft64_10.dll-03-06 22:51:53.101778: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library curand64_10.dll-03-06 22:51:53.109522: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cusolver64_10.dll-03-06 22:51:53.112297: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cusparse64_11.dll-03-06 22:51:53.112870: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudnn64_8.dll-03-06 22:51:53.112999: I tensorflow/core/common_runtime/gpu/:1862] Adding visible gpu devices: 0-03-06 22:51:53.113308: I tensorflow/core/platform/:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.-03-06 22:51:53.114119: I tensorflow/core/common_runtime/gpu/:1720] Found device 0 with properties: pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3070 computeCapability: 8.6coreClock: 1.755GHz coreCount: 46 deviceMemorySize: 8.00GiB deviceMemoryBandwidth: 417.29GiB/s-03-06 22:51:53.114374: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudart64_110.dll-03-06 22:51:53.114483: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublas64_11.dll-03-06 22:51:53.114591: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublasLt64_11.dll-03-06 22:51:53.114692: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cufft64_10.dll-03-06 22:51:53.114785: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library curand64_10.dll-03-06 22:51:53.114881: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cusolver64_10.dll-03-06 22:51:53.114980: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cusparse64_11.dll-03-06 22:51:53.115076: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudnn64_8.dll-03-06 22:51:53.115187: I tensorflow/core/common_runtime/gpu/:1862] Adding visible gpu devices: 0-03-06 22:51:53.585208: I tensorflow/core/common_runtime/gpu/:1261] Device interconnect StreamExecutor with strength 1 edge matrix:-03-06 22:51:53.585330: I tensorflow/core/common_runtime/gpu/:1267]0 -03-06 22:51:53.585394: I tensorflow/core/common_runtime/gpu/:1280] 0: N -03-06 22:51:53.585616: I tensorflow/core/common_runtime/gpu/:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6573 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 3070, pci bus id: 0000:01:00.0, compute capability: 8.6)-03-06 22:51:53.586592: I tensorflow/compiler/jit/:99] Not creating XLA devices, tf_xla_enable_xla_devices not set-03-06 22:51:53.917241: I tensorflow/compiler/mlir/:116] None of the MLIR optimization passes are enabled (registered 2)Epoch 1/7-03-06 22:51:54.744603: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublas64_11.dll-03-06 22:51:55.357995: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cublasLt64_11.dll-03-06 22:51:55.362788: I tensorflow/stream_executor/platform/default/:49] Successfully opened dynamic library cudnn64_8.dll1/40 [..............................] - ETA: 1:33 - loss: 0.7500 - accuracy: 0.3333-03-06 22:51:56.317708: I tensorflow/stream_executor/cuda/:1838] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.40/40 [==============================] - 3s 16ms/step - loss: 0.4230 - accuracy: 0.8329 - val_loss: 0.1822 - val_accuracy: 0.9545Epoch 2/740/40 [==============================] - 0s 7ms/step - loss: 0.3202 - accuracy: 0.9040 - val_loss: 0.1702 - val_accuracy: 0.9545Epoch 3/740/40 [==============================] - 0s 7ms/step - loss: 0.2662 - accuracy: 0.9040 - val_loss: 0.1241 - val_accuracy: 0.9545Epoch 4/740/40 [==============================] - 0s 7ms/step - loss: 0.1704 - accuracy: 0.9089 - val_loss: 0.0482 - val_accuracy: 1.0000Epoch 5/740/40 [==============================] - 0s 7ms/step - loss: 0.0522 - accuracy: 0.9981 - val_loss: 0.0148 - val_accuracy: 1.0000Epoch 6/740/40 [==============================] - 0s 7ms/step - loss: 0.0212 - accuracy: 1.0000 - val_loss: 0.0079 - val_accuracy: 1.0000Epoch 7/740/40 [==============================] - 0s 7ms/step - loss: 0.0123 - accuracy: 1.0000 - val_loss: 0.0048 - val_accuracy: 1.0000D:\Program\CONDA\lib\site-packages\sklearn\metrics\_classification.py:1268: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior._warn_prf(average, modifier, msg_start, len(result))D:\Program\CONDA\lib\site-packages\sklearn\metrics\_classification.py:1268: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior._warn_prf(average, modifier, msg_start, len(result))recall0.5precision0.47959183673469385accuracy0.9591836734693877Epoch 1/620/20 [==============================] - 1s 18ms/step - loss: 0.4540 - accuracy: 0.8943 - val_loss: 0. - val_accuracy: 0.9545Epoch 2/620/20 [==============================] - 0s 10ms/step - loss: 0.3300 - accuracy: 0.9040 - val_loss: 0.1801 - val_accuracy: 0.9545Epoch 3/620/20 [==============================] - 0s 10ms/step - loss: 0.2932 - accuracy: 0.9040 - val_loss: 0.1763 - val_accuracy: 0.9545Epoch 4/620/20 [==============================] - 0s 9ms/step - loss: 0.2894 - accuracy: 0.9040 - val_loss: 0.1689 - val_accuracy: 0.9545Epoch 5/620/20 [==============================] - 0s 9ms/step - loss: 0.2875 - accuracy: 0.9040 - val_loss: 0.1584 - val_accuracy: 0.9545Epoch 6/620/20 [==============================] - 0s 10ms/step - loss: 0.2486 - accuracy: 0.9040 - val_loss: 0.1403 - val_accuracy: 0.9545recall0.5precision0.47959183673469385accuracy0.9591836734693877Epoch 1/618/18 [==============================] - 1s 22ms/step - loss: 0.4610 - accuracy: 0.8135 - val_loss: 0.1842 - val_accuracy: 0.9545Epoch 2/618/18 [==============================] - 0s 10ms/step - loss: 0.3438 - accuracy: 0.9025 - val_loss: 0.1815 - val_accuracy: 0.9545Epoch 3/618/18 [==============================] - 0s 10ms/step - loss: 0.3157 - accuracy: 0.9025 - val_loss: 0.1755 - val_accuracy: 0.9545Epoch 4/618/18 [==============================] - 0s 8ms/step - loss: 0.3025 - accuracy: 0.9025 - val_loss: 0.1599 - val_accuracy: 0.9545Epoch 5/618/18 [==============================] - 0s 8ms/step - loss: 0.2707 - accuracy: 0.9025 - val_loss: 0.1373 - val_accuracy: 0.9545Epoch 6/618/18 [==============================] - 0s 9ms/step - loss: 0.2300 - accuracy: 0.9025 - val_loss: 0.1047 - val_accuracy: 0.9545recall0.5precision0.47959183673469385accuracy0.9591836734693877Epoch 1/5
完整代码和数据链接
