17届华为杯数学建模大赛B题代码
2024-04-25 06:56:29
1 导入库：
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import random
from sklearn.linear_model import Lasso
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor
from sklearn.svm import SVR
from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import ShuffleSplit
from sklearn.feature_selection import RFE
from sklearn.linear_model import LinearRegression
from sklearn.manifold import TSNE
from sklearn.metrics import mean_absolute_error
from sklearn.linear_model import Ridge
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.ensemble import AdaBoostRegressor
from xgboost import XGBRegressor
from sklearn.metrics import mean_squared_error
sns.set_style('darkgrid')
plt.rcParams['font.sans-serif'] = ['SimHei']
2 读取数据：
data1 = pd.read_excel('./附件一.xlsx')
data_285 = pd.read_excel('./285_313.xlsx',sheet_name='Sheet2')
data_313 = pd.read_excel('./313.xlsx')
data4= pd.read_excel('./附件四：354个操作变量信息_改.xlsx')
3 降维：
3.1 用附件四找出存在超出范围的点
standard = []
#将每个特征的取值范围提取出来，存成列表
for each in data4['取值范围']:v = each.split('_')standard.append(v)
standard_val = []
for row in standard:a = []for j in row:a.append(eval(j))standard_val.append(a)
standard_dic = {}
for i,each in enumerate(data4['位号']):standard_dic[each] = standard_val[i]
standard_result = {}
#检测超出或低于特征规定范围的样本
for k,v in standard_dic.items():col_val = data1[k]thre1 = v[0]thre2 = v[1]index = []for inx,j in enumerate(col_val):if j>thre2 or j<thre1:index.append(inx)standard_result[k]=indexbad_features = {}
#提取出超出范围数不为0的特征
for k,v in standard_result.items():if len(v) !=0:bad_features[k]=v#绘制特征超出范围的样本的数量
bad_feature_names = list(bad_features.keys())
bad_feature_vals = []
for i in bad_features.values():n = len(i)bad_feature_vals.append(n)
print(bad_feature_vals)
bad_features_total = pd.DataFrame({'Bad Features':bad_feature_names,'Number of bad features':bad_feature_vals})
bad_features_total_sort = bad_features_total.sort_values(by='Number of bad features',ascending=False)
plt.figure(figsize=(10,8),dpi=100)
sns.barplot('Bad Features','Number of bad features',data=bad_features_total_sort)
plt.xticks(rotation=90)
# plt.title('异常值占比',fontsize=16,fontweight='bold')
plt.xlabel('特征名',fontsize=14,fontweight='bold')
plt.ylabel('异常值数',fontsize=14,fontweight='bold')
plt.savefig('./异常值占比.jpg')
plt.show()
X = data1_copy.drop('RON损失\n（不是变量）',axis=1)
3.2 找出含0值较多的特征
def missing_data(data):"""将原始数据集中为0的值全部转为nanInput:data:原始数据return:data_:缺失值转化后的数据集"""columns = list(data.columns)index_list={}for each in columns:index=[]col = data[each]for inx,v in enumerate(col):if v == 0:index.append(inx)index_list[each]=indexfinal_index = {}for key in index_list.keys():if len(index_list[key]) != 0:final_index[key] = index_list[key]data_ = datafor each in final_index.keys():data_[each].iloc[final_index[each]] = np.nanreturn data_if __name__ == "__main__":Data = missing_data(data1)print('Data Size:{}'.format(Data.shape))print('----------------------------------------------------------------------------------------------')print('Missing proportion:\n',Data.isnull().mean().sort_values(ascending=False).head(33))Missing_proportion = pd.DataFrame({'Proportion':Data.isnull().mean().sort_values(ascending=False).head(32)})plt.figure(figsize=(12,8),dpi=100)plt.rcParams['font.sans-serif'] = ['SimHei']sns.barplot(Missing_proportion.index,Missing_proportion.Proportion)# plt.title('含零特征中的零值占比',fontsize=16,fontweight='bold')plt.xlabel('特征名',fontsize=12,fontweight='bold')plt.ylabel('比例',fontsize=12,fontweight='bold')plt.xticks(rotation=90,fontsize=8)plt.savefig('./Missing proportion.jpg')plt.show()
3.3 删除含10%以上零值的特征和含5个异常值以上的特征
bad_feature_names = list(bad_features.keys())
bad_feature_vals = []
for i in bad_features.values():n = len(i)bad_feature_vals.append(n)bad_features_total = pd.DataFrame({'Bad Features':bad_feature_names,'Number of bad features':bad_feature_vals})
bad_features_total_sort = bad_features_total.sort_values(by='Number of bad features',ascending=False)missing_name = list(Missing_proportion.index)
missing_standard = []
for each in missing_name:thre = standard_dic[each]thre1 = thre[0]thre2 = thre[1]if (thre1<0 and thre2>=0) or (thre1==0):missing_standard.append(each)
miss_list = list(Missing_proportion.loc[Missing_proportion['Proportion']>0.1].index)# 删除0值占10%以上的特征
outlinear = list(bad_features_total_sort.loc[bad_features_total_sort['Number of bad features']>5]['Bad Features'])
# 删除按工艺标准检测出来包含5个样本以上的特征delete_feature = set([])
# 提取要删除的特征
for m in miss_list:delete_feature.add(m)
for o in outlinear:delete_feature.add(o)
X_drop = X.drop(delete_feature,axis=1)
3.4 用RFR+REF选择35个特征
#先对数据进行归一化
X_drop_norm = (X_drop-X_drop.mean())/X_drop.std()
X_train,X_test,y_train,y_test = train_test_split(X_drop_norm,Y,test_size=0.3)
RF = RandomForestRegressor(random_state=0,n_jobs=-1)
rfe = RFE(RF,n_features_to_select=35,step=1)
rfe.fit(X_train,y_train)
features_35 = list(X_train.columns[rfe.support_])
X_norm_35 = X_drop_norm[features_35]
3.5 用Pearson相关性检验删除与过多特征有高度相关性的特征
corr = {}
X_index = list(X_norm_35_corr.index)
for inx,i in enumerate(list(X_norm_35_corr.columns)):corr_list = []val = X_norm_35_corr[i]feature = X_index[inx]for j,each in enumerate(val):if inx!=j and each>0.5:corr_list.append(list(X_norm_35_corr.columns)[j])corr[i] = corr_listfinal_corr = {}
for k,v in corr.items():if len(v)>0:final_corr[k]=v
num_dict = {}
for k,v in final_corr.items():n = len(v)num_dict[k] = n
sorted_num_dict = sorted(num_dict.items(),key=lambda x:x[1],reverse=True)
delete_features = ['辛烷值RON.1','S-ZORB.AT-0009.DACA.PV','S-ZORB.PT_7107.DACA','S-ZORB.PT_7103.DACA','S-ZORB.TC_2801.PV','S-ZORB.FC_2801.PV']
X_norm_final = X_norm_35.drop(delete_features,axis=1)
plt.figure(figsize=(24,24))
sns.heatmap(X_norm_final.corr(),square=True,linewidth=4,linecolor='black',annot_kws={'size':12})
plt.savefig('./最后的特征相关性.jpg')
plt.show()
4.第三问建模
X_train,X_test,y_train,y_test = train_test_split(X_norm_final,Y,test_size=0.2,shuffle=True)
mae_score = {}
mse_score = {}
R2_score = {}#线性回归
lr = LinearRegression()
lr.fit(X_train,y_train)
lr_pred = lr.predict(X_test)
mae_score['LR'] = mean_absolute_error(y_test,lr_pred)
mse_score['LR'] = mean_squared_error(y_test,lr_pred)
R2_score['LR'] = r2_score(y_test,lr_pred)#随机森林
rf = RandomForestRegressor(random_state=5)
rf.fit(X_train,y_train)
rf_pred = rf.predict(X_test)
mae_score['RF'] = mean_absolute_error(y_test,rf_pred)
mse_score['RF'] = mean_squared_error(y_test,rf_pred)
R2_score['RF'] = r2_score(y_test,rf_pred)#LASSO:
LS = Lasso(alpha=0.0005,random_state=5)
LS.fit(X_train,y_train)
LS_pred = LS.predict(X_test)
mae_score['LS'] = mean_absolute_error(y_test,LS_pred)
mse_score['LS'] = mean_squared_error(y_test,LS_pred)
R2_score['LS'] = r2_score(y_test,LS_pred)#SVR:
svr = SVR()
svr.fit(X_train,y_train)
svr_pred = svr.predict(X_test)
mae_score['SVR'] = mean_absolute_error(y_test,svr_pred)
mse_score['SVR'] = mean_squared_error(y_test,svr_pred)
R2_score['SVR'] = r2_score(y_test,svr_pred)#Ridge
ridge =Ridge(alpha=0.002,random_state=5)
ridge.fit(X_train,y_train)
ridge_pred = ridge.predict(X_test)
mae_score['Ridge'] = mean_absolute_error(y_test,ridge_pred)
mse_score['Ridge'] = mean_squared_error(y_test,ridge_pred)
R2_score['Ridge'] = r2_score(y_test,ridge_pred)#Gradient Boosting Regression
GBR =GradientBoostingRegressor(n_estimators=300, learning_rate=0.05, max_depth=4, random_state=5)
GBR.fit(X_train,y_train)
GBR_pred = GBR.predict(X_test)
mae_score['GBR'] = mean_absolute_error(y_test,GBR_pred)
mse_score['GBR'] = mean_squared_error(y_test,GBR_pred)
R2_score['GBR'] = r2_score(y_test,GBR_pred)print('MAE:-------------------------')
print(mae_score)
print('MSE:-------------------------')
print(mse_score)
print('R2:--------------------------')
print(R2_score)
4.1 10-折交叉验证
n_folds = 10
cross_score = {}
scores = cross_val_score(lr, X_norm_final, Y, scoring='neg_mean_squared_error', cv=n_folds)
lr_mae_scores = np.sqrt(-scores)
cross_score['LinearRegression'] =lr_mae_scores.mean().round(decimals=3)
print('For LR model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(lr_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(lr_mae_scores.std().round(decimals=3)))scores = cross_val_score(rf, X_norm_final, Y, scoring='neg_mean_squared_error', cv=n_folds)
rf_mae_scores = np.sqrt(-scores)
cross_score['RandomForest'] =rf_mae_scores.mean().round(decimals=3)
print('For RF model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(rf_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(rf_mae_scores.std().round(decimals=3)))scores = cross_val_score(LS, X_norm_final, Y , scoring='neg_mean_squared_error', cv=n_folds)
ls_mae_scores = np.sqrt(-scores)
cross_score['Lasso'] =ls_mae_scores.mean().round(decimals=3)
print('For LS model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(ls_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(ls_mae_scores.std().round(decimals=3)))scores = cross_val_score(svr,X_norm_final, Y , scoring='neg_mean_squared_error', cv=n_folds)
svr_mae_scores = np.sqrt(-scores)
cross_score['SVR'] =svr_mae_scores.mean().round(decimals=3)
print('For svr model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(svr_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(svr_mae_scores.std().round(decimals=3)))scores = cross_val_score(ridge,X_norm_final, Y , scoring='neg_mean_squared_error', cv=n_folds)
ridge_mae_scores = np.sqrt(-scores)
cross_score['Ridge'] =ridge_mae_scores.mean().round(decimals=3)
print('For ridge model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(ridge_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(ridge_mae_scores.std().round(decimals=3)))scores = cross_val_score(GBR, X_norm_final, Y , scoring='neg_mean_squared_error', cv=n_folds)
gbr_mae_scores = np.sqrt(-scores)
cross_score['Gradient Boosting Regression'] =gbr_mae_scores.mean().round(decimals=3)
print('For GBR model:')
# print(lasso_mae_scores.round(decimals=2))
print('Mean RMSE = ' + str(gbr_mae_scores.mean().round(decimals=3)))
print('Error std deviation = ' +str(gbr_mae_scores.std().round(decimals=3)))model_names = list(cross_score.keys())
model_RMSE = list(cross_score.values())
plt.figure(figsize=(12,10))
sns.barplot(model_names,model_RMSE)
plt.xlabel('模型名称',fontsize=14,fontweight='bold')
plt.ylabel('RMES',fontsize=14,fontweight='bold')
plt.savefig('D:/研究生文献/17届研究生数学建模大赛/2020年中国研究生数学建模竞赛赛题/2020年B题/数模题/图/交叉验证中各个模型的平均误差.jpg')
plt.show()
4.2 调参
rmse_list = []
for i in range(100,1000,50):final_rf = RandomForestRegressor(n_estimators=i,oob_score=True,random_state=5)final_rf.fit(X_train,y_train)scores = cross_val_score(final_rf, X_norm_final, Y, scoring='neg_mean_squared_error', cv=n_folds)final_rf_mae_scores = np.sqrt(-scores)rmse = final_rf_mae_scores.mean().round(decimals=3)rmse_list.append(rmse)
plt.figure(figsize=(14,12))
sns.relplot(np.arange(100,1000,50),rmse_list,kind='line')
plt.xlabel('决策树个数',fontsize=14,fontweight='bold')
plt.ylabel('RMSE',fontsize=14,fontweight='bold')
plt.savefig('./不同数量树情况下的RMSE.jpg')
plt.show()
4.3 第三问最终模型：
final_rf = RandomForestRegressor(n_estimators=550,oob_score=True,random_state=5)
final_rf.fit(X_train,y_train)
5.1 第四问建模：
Y_2 = data1['硫含量,μg/g.1']
X_train_2,X_test_2,y_train_2,y_test_2 = train_test_split(X_norm_final,Y_2,test_size=0.2,shuffle=True)
mae_score_2 = {}
mse_score_2 = {}
R2_score_2 = {}#线性回归
lr_2 = LinearRegression()
lr_2.fit(X_train_2,y_train_2)
lr_pred_2 = lr_2.predict(X_test_2)
mae_score_2['LR'] = mean_absolute_error(y_test_2,lr_pred_2)
mse_score_2['LR'] = mean_squared_error(y_test_2,lr_pred_2)
R2_score_2['LR'] = r2_score(y_test_2,lr_pred_2)#随机森林
rf_2 = RandomForestRegressor(random_state=5)
rf_2.fit(X_train_2,y_train_2)
rf_pred_2 = rf_2.predict(X_test_2)
mae_score_2['RF'] = mean_absolute_error(y_test_2,rf_pred_2)
mse_score_2['RF'] = mean_squared_error(y_test_2,rf_pred_2)
R2_score_2['RF'] = r2_score(y_test_2,rf_pred_2)#LASSO:
LS_2 = Lasso(alpha=0.0005,random_state=5)
LS_2.fit(X_train_2,y_train_2)
LS_pred_2 = LS_2.predict(X_test_2)
mae_score_2['LS'] = mean_absolute_error(y_test_2,LS_pred_2)
mse_score_2['LS'] = mean_squared_error(y_test_2,LS_pred_2)
R2_score_2['LS'] = r2_score(y_test_2,LS_pred_2)#SVR:
svr_2 = SVR()
svr_2.fit(X_train_2,y_train_2)
svr_pred_2 = svr_2.predict(X_test_2)
mae_score_2['SVR'] = mean_absolute_error(y_test_2,svr_pred_2)
mse_score_2['SVR'] = mean_squared_error(y_test_2,svr_pred_2)
R2_score_2['SVR'] = r2_score(y_test_2,svr_pred_2)#Ridge
ridge_2 =Ridge(alpha=0.002,random_state=5)
ridge_2.fit(X_train_2,y_train_2)
ridge_pred_2 = ridge_2.predict(X_test_2)
mae_score_2['Ridge'] = mean_absolute_error(y_test_2,ridge_pred_2)
mse_score_2['Ridge'] = mean_squared_error(y_test_2,ridge_pred_2)
R2_score_2['Ridge'] = r2_score(y_test_2,ridge_pred_2)#Gradient Boosting Regression
GBR_2 =GradientBoostingRegressor(n_estimators=300, learning_rate=0.05, max_depth=4, random_state=5)
GBR_2.fit(X_train_2,y_train_2)
GBR_pred_2 = GBR_2.predict(X_test_2)
mae_score_2['GBR'] = mean_absolute_error(y_test_2,GBR_pred_2)
mse_score_2['GBR'] = mean_squared_error(y_test_2,GBR_pred_2)
R2_score_2['GBR'] = r2_score(y_test_2,GBR_pred_2)print('MAE:-------------------------')
print(mae_score_2)
print('MSE:-------------------------')
print(mse_score_2)
print('R2:--------------------------')
print(R2_score_2)
5.2 问题四模型调参
C = [0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1,1.1,1.2,1.3,1.4,1.5,1.6,1.7,1.8,2]
mse_list = []
for c in C:svr = SVR(kernel='rbf',C=c)svr.fit(X_train_2,y_train_2)pred = svr.predict(X_test_2)mse_list.append(mean_squared_error(y_test_2,pred))
plt.plot(C,mse_list)
plt.show()
5.3 问题四最终模型
final_svr = SVR(kernel='rbf',C=1.75)
final_svr.fit(X_train_2,y_train_2)
5.4 问题四优化建模
5.4.1 根据附件四找出主要操作变量的取值
step_list = []
object_standard = {}
for inx,each in enumerate(data4['位号']):if each in list(X_norm_final.columns)[2:]:object_standard[each] = standard_dic[each]step = data4['Δ值'][inx]step_list.append(step)object_region = {}
count = 0
for each,v in list(object_standard.items()):b = []if step_list[count]>0:step = int(step_list[count]*100)thre1 = int(v[0]*100)thre2 = int(v[1]*100)b.append(thre1-step)a = [i for i in range(thre1,thre2,step)]a.append(thre2+step)for j in a:b.append(j)object_region[each] = np.array(b)/100count+=1elif step_list[count]<0:b = [-0.6,-0.5,-0.4,-0.3,-0.2,-0.15,-0.05]object_region[each] = np.array(b)count+=1
5.4.2 完全随机优化
def cal_dp(x,y,model):pred = model.predict(x)drop = (y - pred)/yreturn drop
def cal_sul(x,model):return model.predict(x)def get_random(object_val):r = []for k,v in object_val.items():d = random.choice(v)r.append(d)return r
def get_random_v(object_val,index,inx):v= list(object_val.values())[index]return v[inx]def random_optimize(Y,index,model1,model2,original_data,object_val):x = original_data[index].reshape(1,-1)y = Y[index]original_norm = (original_data - original_data.mean())/original_data.std()x_norm = original_norm[index].reshape(1,-1)d_p = cal_dp(x_norm,y,model1)sul = cal_sul(x_norm,model2)while True:if d_p<0.3 or sul>5:r = get_random(object_val)original_data[index][2:] = np.array(r)x = original_data[index].reshape(1,-1)original_norm = (original_data - original_data.mean())/original_data.std()x_norm = original_norm[index].reshape(1,-1)d_p = cal_dp(x_norm,y,model1)sul = cal_sul(x_norm,model2)else:print('got!')return d_p,sul,xbreakif __name__ == '__main__':def cal_dp(x,y,model):pred = model.predict(x)drop = (y - pred)/yreturn dropdef cal_sul(x,model):return model.predict(x)def get_random(object_val):r = []for k,v in object_val.items():d = random.choice(v)r.append(d)return rdef get_random_v(object_val,index,inx):v= list(object_val.values())[index]return v[inx]def random_optimize(Y,index,model1,model2,original_data,object_val):x = original_data[index].reshape(1,-1)y = Y[index]original_norm = (original_data - original_data.mean())/original_data.std()x_norm = original_norm[index].reshape(1,-1)d_p = cal_dp(x_norm,y,model1)sul = cal_sul(x_norm,model2)while True:if d_p<0.2 or sul>5:r = get_random(object_val)original_data[index][2:] = np.array(r)x = original_data[index].reshape(1,-1)original_norm = (original_data - original_data.mean())/original_data.std()x_norm = original_norm[index].reshape(1,-1)d_p = cal_dp(x_norm,y,model1)sul = cal_sul(x_norm,model2)else:print('got!')return d_p,sul,xbreakif __name__ == '__main__':d_p_list = []sul_list = []x_list = []for i in range(len(original_data)):d_p,sul,x = random_optimize(np.array(Y2),i,final_rf,final_svr,np.array(original_data),object_region)d_p_list.append(d_p)sul_list.append(sul)x_list.append(x)
17届华为杯数学建模大赛B题代码相关推荐
- 2019年华为杯数学建模大赛D题画汽车行驶路径
程序需要用到python的四个库,分别是folium.numpy.xlrd和os.我用的是python2.7版本,只有os是自带的,剩下三个库都需要用pip安装.程序的大致思路就是,先读入经纬度的ex ...
- 【比赛总结】从编程位队长的角度看待第十三届华中杯数学建模比赛A题
前言 有幸以编程位和队长的身份大一就参加了一次数学建模比赛,这次比赛是"华中杯",所以第一次打还是比较有新鲜感和有很多收获的,故记于此. 因为--在前期找指导老师的时候一说是大一的 ...
- 【2022年华为杯数学建模】B题 方形件组批优化问题 方案及MATLAB代码实现（更新完毕）
更新时间 (1)2022-10-6,13:00 更新内容:二维举行下料求解算法的相关参考文献 (2)2022-10-6,17:55 更新内容:问题一B题建立的数学模型和求解方案 (3)2022-10- ...
- 【自我解析】2020华为杯数学建模比赛C题
C题相较于A题专业性不是那么强,更偏向对给出数据的处理.这里我看了三篇优秀论文进行此题的学习,分别为0319(同济).0034(上海理工).0095(上海交通). 题目提供两个数据附件.P300的介绍 ...
- 【数学建模】2023年华为杯第十九届研究生数学建模大赛E题
2023年华为杯第十九届研究生数学建模大赛E题 <4天半时间><国二> 开源:TODO 放羊的题目,6个问题,当时脑子都想麻了 开个坑,后面再写吧
- 华为杯数学建模思路整理
华为杯数学建模思路整理 1.对于题目: 1.1 题目加上使用的模型进行命名,一定要注意字眼和格式:1.2 目录格式,简洁清爽,表现出使用的模型:1.3 别人的东西,加上文献引用;(**)1.4 对于评 ...
- 2018年华为杯数学建模获奖名单挖掘分析
2018年"华为杯"数学建模获奖名单挖掘分析 本文针对2018年"华为杯"数学建模6张拟获奖名单做一次简单的统计分析,以下分析结果仅代表思路,勿要以假乱真,权威 ...
- 【2022年华为杯数学建模E题赛后总结加思路详细介绍配代码----10月11号写的总结】
提示:下文将介绍2022年华为杯数学建模E题赛后总结加思路详细介绍配代码 傻逼队友,傻逼队友,傻逼队友一定要看好人在进行组队,这是劝告. 这里有几点总结进行描述: 第一,图一定要尽量多,对图的解释要多 ...
- 2022华为杯数学建模B题——方形件组批优化问题
本文章用于记录2022华为杯数学建模B题--方形件组批优化问题 提醒:大家最终提交的时候记得是pdf文件,不是word文件. 第一次参加建模比赛,经验不足,准备不充分,能力也欠缺.考虑2023年继续再 ...
最新文章
- 某程序员为让公司裁掉自己，消极怠工！故意旷工！但公司坚持不裁他，领导：给你发工资，就是不裁你！...
- Debian下使用OpenLDAP 管理端
- Android界面开发问题总结
- 创纪录!Oracle关键补丁更新修复关键漏洞曝光
- 购物商城Web开发第二十三天
- python json解析 超过对象_json序列化数据超出最大值(maxJsonLength)
- LeetCode 其他部分 简单 Python实现
- Clojure 学习入门（3）- 数字类型
- MySQL数据库开发的三十六条军规
- java 输入语句_java怎样输入语句
- X1000对于CPU Core的参数解读（MIPS Cache）
- Tmux(-yank,-cssh,-xpanes)使用指南
- Url scheme 方式 打开App Store和应用商店
- Appium+Pytest+Allure集成PO项目管理模式实现自动化测试
- unity手游之聊天SDK集成与使用二
- VR全景拍摄怎么设置相机
- g729源码分析-5-基音周期搜索(下)
- LeetCode——二叉树
- 古龙妙语大全 下 朱近墨
- Win32 API 学习（20221010）