先做个广告:如需购买GPT帐号或代充值Chatgpt Plus(GPT4)会员,请添加站长客服微信:GPTchongzhi(备注好你的需求)
推荐使用GPT中文版,国内可直接访问:https://ai.gpt86.top
阅读本文大概需要 7 分钟。
来自:blog.csdn.net/weixin_42475060/article/details/129399125
语法更正 文本翻译 语言转换 代码解释-1 代码解释-2 修复代码错误 作为百科全书 信息提取 好友聊天 创意生成器 采访问题 论文大纲 故事创作 问题类比 创建SQL需求 情感分析 将产品描述转变为广告 关键字提取 闲聊机器人
语法更正
文本翻译
语言转换
import numpy as np
import plotly.express as px
def thresholding_algo(y, lag, threshold):
"""
:param y: 输入列表
:param lag: 滑动窗口大小
:param threshold: 调节系数,用于调节容忍范围的大小。
:return:
"""
# signals:信号列表
signals = np.zeros(len(y))
avgFilter = np.zeros(len(y))
# 初始化平均值列表、差值列表
avgFilter[lag - 1] = np.mean(y[0:lag])
dif_li = [0] * len(y)
for i in range(lag, len(y)):
if i < len(y) - lag - 30:
for j in range(i + 11, len(y)):
if y[j] >= y[i - 1]:
break
if j >= len(y) - 30:
back = y[i - 1]
else:
back = np.mean(y[j + 11:j + 20 + lag])
else:
back = y[i - 1]
# 前后均值差值计算
tmp = abs(back - avgFilter[i - 1])
dif = (tmp if tmp > 1 else 1)
if abs(y[i] - avgFilter[i - 1]) > dif * threshold:
signals[i] = (1 if y[i] > avgFilter[i - 1] else -1)
avgFilter[i] = np.mean(y[(i - lag):i])
y[i] = avgFilter[i - 1]
else:
signals[i] = 0
avgFilter[i] = np.mean(y[(i - lag):i])
dif_li[i] = dif
return dict(signals=np.asarray(signals),
avgFilter=np.asarray(avgFilter),
y=np.asarray(y),
dif=np.asarray(dif_li),
)
public static Map
thresholding_algo(double[] y, int lag, double threshold) {
double[] signals = new double[y.length];
double[] avgFilter = new double[y.length];
Arrays.fill(avgFilter, 0.0);
avgFilter[lag - 1] = mean(Arrays.copyOfRange(y, 0, lag));
double[] dif_li = new double[y.length];
Arrays.fill(dif_li, 0.0);
for (int i = lag; i < y.length; i++) {
int j;
if (i < y.length - lag - 30) {
for (j = i + 11; j < y.length; j++) {
if (y[j] >= y[i - 1]) {
break;
}
}
double back;
if (j >= y.length - 30) {
back = y[i - 1];
} else {
back = mean(Arrays.copyOfRange(y, j + 11, j + 20 + lag));
}
} else {
back = y[i - 1];
}
double tmp = Math.abs(back - avgFilter[i - 1]);
double dif = tmp > 1 ? tmp : 1;
if (Math.abs(y[i] - avgFilter[i - 1]) > dif * threshold) {
signals[i] = y[i] > avgFilter[i - 1] ? 1 : -1;
avgFilter[i] = mean(Arrays.copyOfRange(y, i - lag, i));
y[i] = avgFilter[i - 1];
} else {
signals[i] = 0;
avgFilter[i] = mean(Arrays.copyOfRange(y, i - lag, i));
}
dif_li[i] = dif;
}
Mapresult = new HashMap<>();
result.put("signals", signals);
result.put("avgFilter", avgFilter);
result.put("y", y);
result.put("dif", dif_li);
return result;
}
private static double mean(double[] array) {
double sum = 0.0;
for (double d : array) {
sum += d;
}
return sum / array.length;
}
代码解释-1
def Fusion_algorithm(y_list):
"""
最终的融合算法
1、第一次遍历列表: 处理掉小于上一个值的点,使其等于上一个值。
2、第二次使用z-score来处理异常点:一种基于统计方法的时序异常检测算法借鉴了一些经典的统计方法,比如Z-score和移动平均线
该算法将时间序列中的每个数据点都看作是来自一个正态分布,通过计算每个数据点与其临接数据点的平均值和标准差,可以获得Z-score
并将其用于检测异常值,将z-score大于3的数据点视为异常值,缺点:如果异常点太多,则该算法的准确性较差。
3、
:param y_list: 传入需要处理的时间序列
:return:
"""
# 第一次处理
for i in range(1, len(y_list)):
difference = y_list[i] - y_list[i - 1]
if difference <= 0:
y_list[i] = y_list[i - 1]
# 基于突变检测的方法:如果一个数据点的值与前一个数据点的值之间的差异超过某个阈值,
# 则该数据点可能是一个突变的异常点。这种方法需要使用一些突变检测算法,如Z-score突变检测、CUSUM(Cumulative Sum)
# else:
# if abs(difference) > 2 * np.mean(y_list[:i]):
# y_list[i] = y_list[i - 1]
# 第二次处理
# 计算每个点的移动平均值和标准差
ma = np.mean(y_list)
# std = np.std(np.array(y_list))
std = np.std(y_list)
# 计算Z-score
z_score = [(x - ma) / std for x in y_list]
# 检测异常值
for i in range(len(y_list)):
# 如果z-score大于3,则为异常点,去除
if z_score[i] > 3:
print(y_list[i])
y_list[i] = y_list[i - 1]
return y_list
代码解释-2
import numpy as np
from sklearn.ensemble import IsolationForest
import plotly.express as px
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import json
def Fusion_algorithm(y_list):
for i in range(1, len(y_list)):
difference = y_list[i] - y_list[i - 1]
if difference <= 0:
y_list[i] = y_list[i - 1]
# else:
# if abs(difference) > 2 * np.mean(y_list[:i]):
# y_list[i] = y_list[i - 1]
ma = np.mean(y_list)
std = np.std(y_list)
z_score = [(x - ma) / std for x in y_list]
for i in range(len(y_list)):
if z_score[i] > 3:
print(y_list[i])
y_list[i] = y_list[i - 1]
return y_list
修复代码错误
# Buggy Python
import Random
a = random.randint(1,12)
b = random.randint(1,12)
for i in range(10):
question = "What is "+a+" x "+b+"? "
answer = input(question)
if answer = a*b
print (Well done!)
else:
print("No.")
作为百科全书
信息提取
好友聊天
创意生成器
采访问题
论文大纲
class PBA(nn.Module):
def __init__(self, PerformanceThreshold, DistributionType, AttentionWeightRange):
super(PBA, self).__init__()
self.PerformanceThreshold = PerformanceThreshold
self.DistributionType = DistributionType
self.AttentionWeightRange = AttentionWeightRange
def forward(self, input, performance_scores):
# 计算注意力分数
attention_scores = []
for i in range(len(input)):
if performance_scores[i] > self.PerformanceThreshold:
attention_scores.append(performance_scores[i])
else:
attention_scores.append(0.0)
# 将性能分数映射到注意力权重
if self.DistributionType == "softmax":
attention_weights = F.softmax(torch.tensor(attention_scores), dim=0)
elif self.DistributionType == "sigmoid":
attention_weights = torch.sigmoid(torch.tensor(attention_scores))
else:
raise ValueError("Unknown distribution type: {}".format(self.DistributionType))
# 缩放注意力权重到指定范围
attention_weights = attention_weights * (self.AttentionWeightRange[1] - self.AttentionWeightRange[0]) + self.AttentionWeightRange[0]
# 计算加权输入
weighted_input = torch.mul(input, attention_weights.unsqueeze(1).expand_as(input))
output = torch.sum(weighted_input, dim=0)
return output
故事创作
问题类比
创建SQL需求
情感分析
将产品描述转变为广告
关键字提取
闲聊机器人
总结
推荐阅读:
互联网初中高级大厂面试题(9个G) 内容包含Java基础、JavaWeb、MySQL性能优化、JVM、锁、百万并发、消息队列、高性能缓存、反射、Spring全家桶原理、微服务、Zookeeper......等技术栈!
⬇戳阅读原文领取! 朕已阅