Anh vu-assignment 3

[aidanvu1992]

No description provided.

[llpk79]

1. Load a dataset from Github (via its RAW URL)

• Nice job here. Not much to get excited about.

2. Load a dataset from your local machine

• Using the colab library, nice.

3. Load a dataset from UCI using !wget
   -Good job here as well. Would be interesting to see where you can take the poker data.

[llpk79]

Sprint challenge code review:

Part 1 - Load and validate the data

• Load the data as a pandas data frame.
  • You've lost a row because the first row is read as the header by defualt.
  • You've added an unnecessary step. df = pd.read_csv(cancer_survival_url, header=['list', 'of', 'columns']) is sufficient.
• Validate that it has the appropriate number of observations (you can check the raw file, and also read the dataset description from UCI).
  • Use df.shape to view number of rows. !cat <file.name> | wc -l to view number of rows in file, or go to data page to confirm expected rows in dataframe.
• Validate that you have no missing values.
  • Use df.isnull().sum() to see an easy to read summation of null values per column.
• Add informative names to the features.
  • Complete.
• The survival variable is encoded as 1 for surviving >5 years and 2 for not - change this to be 0 for not surviving and 1 for surviving >5 years (0/1 is a more traditional encoding of binary variables)
  • Nicely done.
• At the end, print the first five rows of the dataset to demonstrate the above.
  • Complete.

Part 2 - Examine the distribution and relationships of the features

• Explore the data - create at least 2 tables (can be summary statistics or crosstabulations) and 2 plots illustrating the nature of the data.
  • Good job exploring several cross-tabs. Consider binning the 'Number of positive axillary nodes detected'' column as well.
  • Check out pd.qcut()
  • Try using bar graphs instead of line charts when comparing discreet values. Tend to reserve line graphs for time series.

Part 3 - DataFrame Filtering

• Use DataFrame filtering to subset the data into two smaller dataframes. You should make one dataframe for individuals who survived >5 years and a second dataframe for individuals who did not.
  • It would be more informative to cross-tab with age vs nodes or year vs 'nodesrather than with survival because we know all are eithersurvivedornot_survived`.
• Create a graph with each of the dataframes (can be the same graph type) to show the differences in Age and Number of Positive Axillary Nodes Detected between the two groups.
  • Try bar graphs and heat-maps with cross-tabs as above. Try plotting on the same bar graph with similar cross-tabs for both 'survived' and 'not_survived' populations.

Part 4 - Analysis and Interpretation

• What is at least one feature that looks to have a positive relationship with survival? (As that feature goes up in value rate of survival increases)
  • 👎
  • year_of_operation and survival have a positive relationship.
• What is at least one feature that looks to have a negative relationship with survival? (As that feature goes down in value rate of survival increases)
  • 👍
  • Age and number of nodes are negatively correlated.
• How are those two features related with each other, and what might that mean?
  • 👍
  • Age and year_of_operation are positively correlated.

Not bad, Anh. I'm going to give a 2. Do go back and review some of this material as we will continue to build on this to rapidly more complex applications.