Name: 學習用於數據分析的 Pandas？從這裡開始 (Learning Pandas for Data Analysis? Start Here.)
Uploaded: 2025-05-26T18:37:41.000Z
Duration: 22 min 50 s

In this video, you're going to learn all the essential tricks to using pandas to do data manipulation and wrangling.

在本視頻中，您將學習使用 pandas 進行數據操作和處理的所有基本技巧。

Pandas is one of the most popular Python libraries and the tool for doing data manipulation.

Pandas 是最流行的 Python 庫之一，也是進行數據處理的工具。

情態副詞

And now with the ability to write Python code directly in Excel, pandas is reaching a whole new audience.

現在，有了直接在 Excel 中編寫 Python 代碼的功能，pandas 正在接觸到全新的閱聽人。

With that comes a lot of hidden functionality that you wouldn't necessarily know about without fully diving into the documentation.

隨之而來的是許多隱藏功能，如果不深入瞭解文檔，你不一定會知道這些功能。

That's why I've created this video using my years of experience with this library to boil it down to just the functions that you need to know to get started.

是以，我利用自己多年使用該庫的經驗製作了這段視頻，將其歸納為入門所需的功能。

So if you're a seasoned veteran or if you're just learning pandas for the first time, I suggest you watch this video all the way through and then go back and rewatch any parts that weren't clear to you the first time.

是以，如果你是一個經驗豐富的老手，或者你只是第一次學習熊貓，我建議你把這個視頻看完，然後再回頭重看第一次沒看明白的部分。

If you learn something new, make sure you like this video, write any comments down below and subscribe to the channel for more content.

如果你學到了新知識，請確保你喜歡這段視頻，在下面寫下任何評論，並訂閱該頻道以獲取更多內容。

Of course, the first thing we need to do is import pandas in order to work with it.

當然，我們首先要做的是導入 pandas，以便使用它。

I'm also printing out here my system info and the pandas version I'm using, which is the latest at the time of recording this video.

我還在這裡打印出了我的系統資訊和我使用的 pandas 版本，這是錄製這段視頻時的最新版本。

In most of my videos, I get at least three or four comments asking about the interpreter that I'm using.

在我的大部分視頻中，至少會有三四條評論詢問我使用的翻譯器。

This IDE is called JupyterLab and I have a solarized dark theme on it.

這個 IDE 叫做 JupyterLab，我在上面安裝了一個日光暗色主題。

But I talk all about Jupyter notebooks and JupyterLab in a complete video that you should see a link for in the description.

不過，我在一個完整的視頻中講述了 Jupyter 筆記本和 JupyterLab 的全部內容，你應該能在描述中看到相關鏈接。

With pandas, you're typically going to be working with a pandas data frame.

使用 pandas 時，你通常會使用一個 pandas 數據幀。

You can think of a pandas data frame as being similar to a spreadsheet where you have a data set that contains columns and rows.

你可以把 pandas 數據幀想象成類似於電子表格，其中有一個包含列和行的數據集。

While you can make a data frame from scratch using raw data in Python, the majority of the time you're going to be reading from either a file or a database.

雖然您可以使用 Python 中的原始數據從頭開始創建數據幀，但大多數情況下，您需要從文件或數據庫中讀取數據。

Luckily, pandas has a lot of different methods for reading in different file types.

幸運的是，pandas 有很多不同的方法來讀取不同類型的文件。

All of these methods start with read and they include reading from a CSV file, Excel files, and many other file formats.

所有這些方法都以讀取開始，包括從 CSV 文件、Excel 文件和許多其他文件格式中讀取。

Each read method comes with its own parameters that can be set when reading in the files.

每種讀取方法都有自己的參數，可以在讀取文件時設置。

Typically, the default settings should work pretty well, but especially with read CSV, which is one of the most popular file formats, you have a lot of parameters to choose from.

通常情況下，默認設置應能很好地工作，但特別是讀取 CSV 這種最常用的文件格式時，你有很多參數可供選擇。

This could almost be an entire video in itself, but some of the main reading parameters you may want to change is changing the delimiter, say if you had a tab-separated file, setting use columns if you only wanted to read in a subset of the columns, automatically parsing dates when reading in the file, and reading in large files in chunks.

這幾乎可以製作一整段視頻，但你可能需要更改的一些主要讀取參數包括：更改分隔符（例如，如果你有一個以製表符分隔的文件）、設置使用列（如果你只想讀取列的子集）、讀取文件時自動解析日期，以及分塊讀取大文件。

Now in this tutorial, we're going to be working almost entirely with this flights dataset, which I have saved as a parquet file.

在本教程中，我們將幾乎完全使用我保存為 parquet 文件的班機數據集。

We'll read this in as df, which stands for data frame.

我們將以 df 的形式讀入，它代表數據幀。

If we do a type on this df, you can see it is a pandas core data frame.

如果我們對這個 df 進行鍵入，可以看到它是一個 pandas 核心數據幀。

And similar to reading, there's a lot of different ways you can write your data frame to files or to databases using the to SQL.

與讀取類似，有很多不同的方法可以將數據幀寫入文件或使用 to SQL 寫入數據庫。

Most of these writing formats typically just take in the file path where you plan to write the file, and each writing method has its own set of parameters you can choose from.

大多數寫入格式通常只需要輸入您計劃寫入文件的文件路徑，每種寫入方法都有自己的一組參數供您選擇。

For instance, when writing to a CSV, I often find myself setting index to faults so that the index is not saved when it doesn't have relevant information.

例如，在寫入 CSV 時，我經常發現自己將索引設置為故障，這樣在沒有相關資訊時就不會保存索引。

Now, once you do have your data frame, there are many techniques that you may want to use to understand what the data looks like.

現在，一旦你有了數據框架，你可能想使用許多技術來了解數據的外觀。

Head lets you see by default the first five rows of your data frame, depending on how many columns you have.

默認情況下，Head 可以讓你看到數據幀的前五行，具體取決於你有多少列。

It may hide the middle columns when displaying this, but you can override this by setting this option to display a larger maximum columns.

在顯示時可能會隱藏中間列，但您可以通過設置該選項來覆蓋這一點，以顯示更大的最大列數。

And you can see here this dataset has a lot of columns which are now displayed.

你可以看到這個數據集有很多列，現在都顯示出來了。

The head command can be given any number to show that number of columns.

標題命令可以給出任何數字，以顯示該列數。

There's also a tail command, which, as you might expect, will show the last five rows in the data frame that sample will pull a random subset of rows from your entire data set.

還有一個 tail 命令，如你所料，它會顯示數據幀中的最後五行，樣本會從整個數據集中隨機抽取一個子集。

You can provide it a number which will return that number of rows or a fraction which will provide that fraction of the entire data set because the results of that sample are random.

您可以向它提供一個數字，它將返回該行數，或者提供一個分數，它將提供整個數據集的該分數，因為該樣本的結果是隨機的。

It can be helpful to add a random state and this will help with reproducibility.

添加隨機狀態可能會有幫助，這將有助於重現性。

Since every data frame consists of columns and rows, we can access the columns directly by using dot columns.

由於每個數據幀都由列和行組成，我們可以使用點列直接訪問列。

This can be helpful if you just want to see a list of all the columns in your data frame.

如果您只想查看數據幀中所有列的列表，這將很有幫助。

Dot index will give you the index values in the data frame.

點索引將顯示數據幀中的索引值。

Data frames have a few methods built in that help you quickly get a top down view of what the data looks like.

數據幀內置了一些方法，可以幫助你快速自頂向下查看數據的外觀。

Dot info will give you information about the data frame, including the size, listing each column and the data type of each column.

Dot info 將提供有關數據框的資訊，包括大小、每列清單和每列的數據類型。

And it's also really helpful at the bottom of dot info.

You see the memory usage of this data set.

你可以看到該數據集的內存使用情況。

For a quicker summary, you can set verbose equals to false in dot info.

如果想更快地獲得摘要，可以在 dot info 中將 verbose 設置為 false。

The built in dot describe function can be run on any data frame to provide some descriptive statistics about the data, including the count values, mean, minimum, some percentiles, maximum and the standard deviation for any numeric column.

內置的 dot describe 函數可在任何數據幀上運行，以提供數據的一些描述性統計，包括任何數值列的計數值、平均值、最小值、某些百分位數、最大值和標準偏差。

Describe can also handle non-numeric columns.

For instance, running on the airline will show instead the count number of unique values, top occurring value and the frequency.

例如，在航空公司上運行時，將顯示唯一值的計數、最高出現值和頻率。

Many times you'll also find yourself wanting to just pull in the shape of the data frame and you can do that by just running dot shape.

很多時候，你也會發現自己想直接調出數據幀的形狀，這可以通過運行點形狀來實現。

Similarly, you can take the length of the data frame, which will give you the same thing as the first value in the shape, which is the number of rows.

同樣，你也可以求出數據幀的長度，這樣得到的結果與形狀中的第一個值（即行數）相同。

Subsetting a data frame is something that you'll often find yourself doing, and you may want to either filter down to a subset of columns or a subset of rows to subset columns.

對數據幀進行子集處理是你經常要做的事情，你可能想要篩選出列的子集，或者篩選出行的子集來子集列。

You can simply provide a list within a bracket at the end of the data frame and that will subset just to those columns.

您只需在數據幀末尾的括號內提供一個列表，就能只對這些列進行子集。

So here we've subset it just to the flight date airline in the origin columns.

是以，在這裡我們只將其子集為原點列中的班機日期航空公司。

And if you remember from before, running DF dot columns will give us a list of the entire column names.

如果你還記得，運行 DF dot columns 可以得到整個列名的列表。

So using this in combination with a slice, we could, for instance, take the first five columns and then subset our entire data frame just to those five or something like this would give us the last five columns.

是以，結合切片使用此功能，例如，我們可以選取前五列，然後將整個數據幀子集為這五列，或者類似於這樣的功能，我們就可以得到最後五列。

And depending on how comfortable you are with list comprehensions, it's pretty easy to filter down the columns based on certain criteria.

根據你對列表理解的熟練程度，很容易就能根據特定條件篩選出列。

For instance, here I'm filtering down just two columns that have the text time in their name, and then I can provide that to the data frame and return only those columns.

例如，在這裡我只篩選出兩列名稱中包含時間文本的列，然後將其提供給數據幀，並只返回這些列。

A less commonly used way of filtering columns is by using the select D types method, which will return a subset of the data frames columns based on the data type you provide it.

一種不太常用的篩選列的方法是使用 select D types 方法，它會根據你提供的數據類型返回數據幀列的子集。

Here you can see we've selected only data types in this data set that are integers.

在這裡，你可以看到我們在這個數據集中只選擇了整數數據類型。

One thing to note about selecting a single column is that if you provide just that column name, you will be returned with the series of that data column as seen here.

選擇單列時需要注意的一點是，如果只提供列名，將返回該數據列的系列，如圖所示。

If instead you prefer to have a data frame with a single column, you need to make sure to include double brackets around the column name.

如果您希望使用單列數據框，則需要確保在列名周圍加上雙括號。

And you can see here this is a data frame to filter a data frame based on its rows.

在這裡，你可以看到這是一個數據框，可以根據行數過濾數據框。

You'll typically be using either ILOC or LOC.

Both LOC and ILOC allow you to access elements in the data frame based on their location.

LOC 和 ILOC 都允許您根據位置訪問數據幀中的元素。

However, ILOC uses the index location and LOC uses the names.

不過，ILOC 使用索引位置，而 LOC 使用名稱。

Because of that, you'll find yourself using LOC a lot more than ILOC.

正因為如此，你會發現自己使用 LOC 比 ILOC 多得多。

To show as an example, if we wanted to use ILOC to pull this value in the data frame, we'd be looking for row number one and column number three.

舉例說明，如果我們想使用 ILOC 在數據幀中提取這個值，我們需要查找第一行和第三列。

So we formulate our location with one being the row and three being the column and we return that value.

是以，我們在設置位置時，1 代表行，3 代表列，然後返回該值。

So if we wanted to pull the first five rows and the first five columns, we could do it like this.

是以，如果我們想提取前五行和前五列，可以這樣做。

Using ILOC this way, we are subsetting both the columns and the rows, but we could also just provide a single value which will only filter to the row.

通過這種方式使用 ILOC，我們對列和行都進行了子集，但我們也可以只提供一個值，這樣就可以只對行進行篩選。

If we put this in a list, then we'll be filtering to row number five as a data frame.

如果我們把它放在一個列表中，那麼我們將以數據幀的形式篩選到第五行。

Similarly, if we wanted to filter only to a column, we can use a colon to state we'd like all rows.

同樣，如果我們只想過濾某一列，我們可以使用冒號來表示我們想要過濾所有行。

And here we're saying we just would like the first column.

在這裡，我們說我們只想要第一列。

Putting this in a list returns the subset as a data frame.

將其放入列表中，子集就會以數據幀的形式返回。

If we wanted to do this exact same subsetting using LOC, instead of using numerical values for their locations, we would actually provide in the name of the columns or rows that we would like to filter to.

如果我們想使用 LOC 進行完全相同的子集設置，就不能使用數值來表示它們的位置，而是要提供我們想要過濾的列或行的名稱。

Here we have the airline and origin column selected.

這裡我們選擇了航空公司和出發地一欄。

One of the most powerful things of using the LOC method is that we're able to filter not only on names, but on Boolean expressions for our rows.

使用 LOC 方法最強大的一點是，我們不僅可以過濾名稱，還可以過濾行的布爾表達式。

How this works is by creating an expression on one or more columns.

具體方法是在一列或多列上創建一個表達式。

然後，我們會返回一個 "真 "或 "假 "的序列。

Here we're looking for where airline equals spirit airlines and then passing this expression into LOC.

在這裡，我們要查找航空公司等於精神航空公司的位置，然後將此表達式傳入 LOC。

Then we'll filter our data frame down to only when this Boolean expression is true.

然後，我們將過濾我們的數據幀，僅當布爾表達式為 "true "時才進行過濾。

As you can see, the result only contains rows with spirit airlines.

如您所見，結果只包含精神航空公司的行。

We can combine Boolean expressions with ands or or statements to filter on multiple things or columns.

我們可以將布爾表達式與 ands 或 or 語句結合起來，對多個事物或列進行篩選。

Here we're filtering on the airline and also to a specific date.

在這裡，我們根據航空公司和具體日期進行篩選。

One helpful tip to be aware of when creating Boolean expressions is that you can easily take the inverse of the expression by simply adding a tilde before it.

在創建布爾表達式時，需要注意的一個有用提示是，只需在表達式前添加一個轉折號，就可以輕鬆地實現表達式的逆轉。

And if we use LOC to filter to this expression, we've now returned all rows except for the rows with spirit airlines and this specific date.

如果我們使用 LOC 對這個表達式進行過濾，那麼除了包含精神航空公司和特定日期的記錄外，我們已經返回了所有記錄。

An alternate way to query your data frame based on Boolean expressions is by using the dot query method.

根據布爾表達式查詢數據幀的另一種方法是使用點查詢方法。

The dot query method takes in a string representation of the Boolean expression you wish to filter on.

點查詢方法接收您要過濾的布爾表達式的字符串表示。

As an example, I can filter for when the departure time is greater than 1130.

例如，我可以篩選出發時間大於 1130 的班機。

Dot query can take in multiple expressions.

As seen here, we're filtering on departure time and the origin.

如圖所示，我們根據出發時間和出發地進行篩選。

Any values in the query method are assumed to be columns, but you can represent strings by wrapping the value in quotes or your query expression can access an external variable by using the at symbol before the name in your query string.

查詢方法中的任何值都被假定為列，但您可以用引號將值包起來來表示字符串，或者在查詢字符串中的名稱前使用 at 符號來訪問外部變量。

Next, let's talk about summarizing data in the data frame.

接下來，我們來談談在數據框中彙總數據的問題。

There are a number of summarization methods that can be run on either a single column or multiple columns.

有許多彙總方法可以在單列或多列上運行。

They include the mean value, the minimum value, the maximum value, standard deviation, variance, count, which will show you the number of values.

它們包括平均值、最小值、最大值、標準偏差、方差、計數（顯示數值的數量）。

You can also sum all the values in the column and dot quantile will give you the quantile values.

您還可以對列中的所有值求和，然後點量值會給出量值。

As you can see here, we have the 25th and 75th quantile for the departure time.

如圖所示，我們有出發時間的第 25 個和第 75 個量值。

Any of these summarization methods can be run on multiple columns.

這些彙總方法中的任何一種都可以在多列上運行。

This will return a pandas series where the index are the column names and the values are the summary values statistics.

這將返回一個 pandas 序列，其中索引是列名，值是彙總值統計量。

字幕列表影片播放

學習用於數據分析的 Pandas？從這裡開始 (Learning Pandas for Data Analysis? Start Here.)

specific

sort

entire

access