官术网_书友最值得收藏!

Exploratory data analysis

For this chapter, we will be using curated data from the New York taxi trip dataset provided by the city of New York. The original source for this data can be found here: https://data.cityofnewyork.us/api/odata/v4/hvrh-b6nb.

Visit the following website for more details about the data that's included in this dataset:   https://data.cityofnewyork.us/Transportation/2016-Green-Taxi-Trip-Data/hvrh-b6nb.

For starters, let's have a peek at the data at hand using pandas. The curated data (NYC_sample.csv) that we will be using here can be found at the following download link: https://drive.google.com/file/d/1OkkYZJEcsdCkU0V42eP6pj6YaK2WCGCE/view.

df = pd.read_csv("NYC_sample.csv")
df.head().T

The curated New York taxi trip data that we are using has around 1.14 million records and has columns related to taxi fare, as well as trip duration, as you can see from the following screenshot: 

The data dictionary for this data that can be found at https://data.cityofnewyork.us/api/views/hvrh-b6nb/files/65544d38-ab44-4187-a789-5701b114a754?download=true&filename=data_dictionary_trip_records_green.pdf. The data download page provides us with useful information about the data.

We can make the following inferences and assumptions about processing the data: 

  • Some columns have missing values (NaN included) that need to be handled.
  • The value that we are trying to predict for each trip is the trip duration. This needs to be derived from the pickup and dropoff times for the training and validation data.
  • Once the trip duration has been derived, we need to get rid of the dropoff time information since our objective is to compute the dropoff time when given other information, such as pickup location and time and dropoff location.
  • We intend to primarily use the pickup and dropoff locations to predict time. Hence, in the training phase, we should be dropping records that don't have pickup and dropoff location information. 
  • We'd like to exclude columns related to trip cost because it doesn't contribute to the model.

We can do the following to tackle the problems we identified by taking a cursory glance at the data:

  • Handle missing values
  • Handle time values
  • Handle unrelated data 
主站蜘蛛池模板: 始兴县| 扎兰屯市| 勃利县| 大关县| 南投县| 屯留县| 鄂托克旗| 双桥区| 长沙市| 蕉岭县| 浮梁县| 无为县| 上饶县| 四子王旗| 定兴县| 呼图壁县| 遵义市| 建平县| 南宁市| 黎城县| 田林县| 新绛县| 香格里拉县| 常州市| 夹江县| 平顺县| 芦溪县| 绿春县| 嘉定区| 旺苍县| 湾仔区| 宝清县| 常宁市| 华坪县| 龙里县| 灵丘县| 桐柏县| 仙桃市| 新蔡县| 太和县| 伊春市|