so just to clarify, UTF8 and Latin1 encoded means UTF8 is used for electronic communication, like the data we process we convert this alpha numeric data into machine language using UTF8 encoding .. and we all know at granular level the machine language is in 0s and 1s, so we need some sort of encoding to convert the incoming English language alphabets into machine language for a machine to process the data ... but when the incoming language has some characters which are not defined in English language then machine will not able to convert it into 0s and 1s cos for it its a foreign word/character not listed in given reference directory which is UTF8 in this case, so we need to provide appropriate reference so in case of incoming data has Japanese character then you can use JIS encoding, in case of incoming data has Latin characters then latin-1 etc ...
Enjoyed going through this EDA session. Sharing my new learnings from my first EDA session as my thanksgiving. 1) Learnt about encoding errors. 2) Learnt about matplotlib figure sizing use rcParams 3) Learnt about why hue did not work in Seaborn as intended. 4) Learnt to understand the difference between value_counts and groupby().size 5) Loved to see how reset_index was being used time and again. 58:08 We use hue when we wish to differentiate the categories. But remember the colours of the bar graph were already in different colours. Why was that? In Seaborn library, the default palette is to show a mix of colours for each bar, which we subsequently changed to the colours we wanted. So the hue parameter did not do anything additional in this context. But the hue parameter did give us a legend. 1:18:45 We used groupby().size plenty of times. I was wondering when to use value_counts and when to use groupby().size df[condition] returns a series ==> Here we can use value_counts(). We cant use groupby when it is a series. df[df[condition]] returns a dataframe ==> We can use groupby().size here. Also size gives total count of all elements. If used with a groupby, it gives count of all elements for each group, including null values. But if we use value_counts, we get the element-wise total excluding null values.
@@iakhileshgupta3553 save the csv file in local folder ... then assign the path example Data= pd.read_csv(C:\\program file \\user\\folder\\'file_name.csv') Then print data.head() ... First 5 columns & rows come into picture ...
@@iakhileshgupta3553 recheck CMD prompt ... whether pandas is installed or not ...if not then ... Pip install Numpy ... Pip install pandas ....both in cmd propmt In Jupyter notebook .. .import Import pandas as pd Import Numpy as np .. THEN RUN IT .. CLICK ON RUN BUTTON ( it has to reset always whenever you open Jupyter notebook .... Then type ... Df= read_csv('dats.csv') df.head ()
Sir @@sidindian1982 i have pandas & numpy installed i have checked on cmd as pip list it is there & also done what you suggested it says already installed but it won't read the file & gives the above permission error in the terminal
on which basis you define the top 10 cuisines..... and one more thing I tried a code final_df[final_df['Rating text']=='Excellent']['Cuisines'].head(10) so I found this code more efficient on the basis of info. which code is more appropricate??
In missing values part the heat map plot is not visible instead if we resize the scale of the graph we will be able to actuallyplot the point. plt.figure(figsize = (15,8)) sns.heatmap(df.isnull(), yticklabels = False, cbar = True, cmap = 'viridis') the figure method will help us to resize the heat map accordingly.
Thank you sir for this! I think, you should conduct more such sessions where you assign us such questions related to Data Analytics and then discuss answers
For the last Assingment which he has given as homework you can replace final_df as what you've coded in your book final_df[["Cuisines"]].groupby(["Cuisines"]).size().reset_index().sort_values(by=0,ascending=False).head(10)
I am from Pakistan and I have seen a lot of videos about EDA, But this one is just amazing. It is really easily understandable for begginers. Respect to you Sir!!!
Hello the explanation is great and i have done the assignment it was super easy: # Finding top 10 cuisines cuisine_counts = df.groupby('Cuisines').size().reset_index() top_10_Cuisines = cuisine_counts.head(10)
Prefect explanation! the visualization, the sub-setting , everything, the queries and the observations! GREAT thanks so much... (i just sttruggle a little bit about when to use groupby)
Hi Krish. Please do an example for the difference between using LSTM for classification and LSTM for regression. Explain the difference between using LSTM for the two. Especially for multivariate. You have always been my teacher. I learned machine learning and deep learning from you. No other bootcamp, I didn't do any computer science course in University. Just your RU-vid videos. Thank you so much.
Hi Krish Amazing video. Thanks alot for all the videos, keep it up 👍 Have a doubt @1.02, why max no of ratings is between 2.5 to 3.4? And not 2.5 to 3.9?
Thanks sir One issue is Krish sir moves his screen rapidly, it is difficult to code along with him, In RU-vid we can pause video and look for that section, but in live class difficult.
i m getting error in barplot where we plotting aggregate rating vs rating count (the error showing value error not able to interpret the rating count) at 55.20 mins
Sir I want to extract features from a text .CSV file, by using TFIDF and Ngram model, and I want the output in sparse matrix, Do you have any tutorial on that plz guide.
Krishna have a lets say confusion rather than doubt as am looking for EDA knowledge if in real time if i was given a data set for doing EDA as am data analyst so my work related to get the insight from the data and to represent to client or have to do Machine learning for predictions also. Please can you give clarity?