Giter Club home page Giter Club logo

weather_prediction's Introduction

weather_prediction

数据集概述

数据集是datasets/weather.csv,该表格包含2013年1月1日-2017年12月31日,来自北京、上海、重庆的共58个站点按天记录的逐日天气观测数据。

  • 形如: date,station,city,county,pressure,wind_speed,wind_direction,temperature,humidity,rain20,rain08,cloud,visibility,phenomenon
    20130101,巴南,重庆市,巴南市,959.1,3.4,999003,6,48,0,0,3,8000,"02,03,"
    20130101,綦江,重庆市,綦江市,962.6,6.5,999012,5.6,59,0,0,35,15000,"02,"
    20130101,渝北,重庆市,渝北市,963.9,4.1,999003,5.9,44,0,0,0,10000,"01,02,03,"
    20130101,梁平,重庆市,梁平市,965.3,2.2,999001,2.1,68,0,0,0,9000,"01,10,02,03,"
    20130101,垫江,重庆市,垫江县,968.9,1.7,999004,3.8,61,0,0,0,8000,"01,02,03,"
    20130101,大足,重庆市,大足市,972.4,2.4,999003,3,72,0,0,0,8000,"01,02,03,10,"
    20130101,长寿,重庆市,长寿市,974,2.3,999001,5.4,61,0,0,0,9000,"10,02,03,"
    20130101,涪陵,重庆市,涪陵市,975,4.1,999003,4.6,70,0,0,3,1200,"10,02,03,"
    20130101,永川,重庆市,永川市,977.2,3.3,999003,5.6,54,0,0,0,7000,"02,03,10,"

  • 字段说明

字段名 中文名称 说明
date 日期
station 站点 观测站点名称
city 城市 站点所属城市
county 市辖区/县/自治县 站点所处县级行政区
pressure 平均气压(百帕)
wind_speed 最大风速(米/秒)
wind_direction 最大风速的风向(度)
temperature 平均气温(摄氏度℃)
humidity 平均相对湿度(百分率)
rain20 20-20时降水量(毫米) 代表观测日期当天20时获取,是从前一日20时至当日20时这24小时的累计降水量数据,是常用的降水量数据
rain08 08-08时降水量(毫米) 前一日8时至观测日8时的累计降水量数据
cloud 平均总云量(百分比) 观测到的天空有效面积内云量的占比
visibility 最小水平能见度(米)
phenomenon 天气现象记录 *见下方“天气现象记录说明”
  • 天气现象记录说明
数值 类别名 说明 对应phenomenon的代码值
0 sunny 晴天/无特殊天气状况 0, 2, 100, 102, 211, 508-511
1 cloud 多云 1, 3, 19, 101, 103
2 rain 降雨 14, 15, 16, 20, 21, 23, 24, 25, 26, 27, 29, 50-69, 80-84, 87-99, 121-123, 125, 126, 140-144, 150-168, 180-184, 190, 192, 193, 195, 196, 227, 250-266, 280-282, 285, 286, 289, 290, 292, 293
3 fog 有雾 4, 10, 11, 12, 28, 40-49, 104, 105,120, 130-135, 224, 241-249
4 haze 有霾 4, 5, 104, 105, 110, 206, 210
5 dust 尘土/沙尘 6, 7, 8, 9, 30-35, 98, 104, 105, 204-209, 220, 221, 230
6 thunder 打雷 17, 29, 91, 95-99, 126, 190-196, 217, 292, 293
7 lightning 闪电 13, 17, 29, 91, 95-99, 112, 126, 190-196, 213, 217, 292, 293
8 snow 降雪 22, 23, 26, 36-39, 68-78, 83-90, 93-95, 97, 111, 122, 124, 127-129, 145-148, 170-173, 177, 178, 185-187, 192, 195, 223, 239, 259, 270-279, 282-287, 290, 291
9 hail 冰雹 23, 27, 79, 87, 89, 93, 94, 96, 99, 174-176, 189, 193, 196, 284-291
10 wind 强风 17, 18, 29, 91, 95-99, 118, 126, 190-196, 199, 217, 292, 293

注意上述各种类别的天气可能会重叠,即在同一天中出现多种天气状况,因此该字段以onehot的形式保存,例如:

sunny cloud rain fog haze dust thunder lightning snow hail wind
0 0 1 0 0 0 0 0 1 1 0

脚本概述

  1. 本项目的主要代码在weather.ipynb中呈现,该脚本主要实现两种任务:回归任务和多标签分类任务。
  2. 回归任务中采用LSTM模型是多对一的形式,即利用处理好的全部26个属性来预测单一属性,如温度,可见度和湿度等。
  3. 多标签分类任务采用LSTM模型是多对多的形式,即利用处理好的全部26个属性来预测气象的11类属性,即上方的天气现象记录说明。
  4. 数据集的划分方式写在get_data()函数中,主要内容是将表格按照时间和县来排列,切分成57个县的天气情况,然后对每个县选择时间窗口滑动进行分组,最后将所有组垂直拼接并shuffle。按照8:1:1的比例划分为训练集、验证集和测试集。该分法的缺陷是不能单独运行测试函数,因为shuffle的时候,将测试集合破坏了,因此只能在训练之后使用测试函数。
  5. 回归任务的准确率使用R2_score来进行评定,预测温度的准确率大约在95%左右。多标签分类任使用f1_score和auc_score来进行评定,最后二者分别为85%和68%,仍然有较大的上升空间。

weather_prediction's People

Contributors

qmh1234567 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.