Mixed NumPy + Pandas Exercises

M1. Convert a DataFrame column to a NumPy array. df['v'].to_numpy()

M2. Apply a NumPy function to a column. np.log1p(df['v'])

M3. Add a column that is the row-wise max of two columns. df['mx']=np.maximum(df.a, df.b)

M4. Standardize all numeric columns. df=(df-df.mean())/df.std()

M5. Flag outliers beyond 3 std. (np.abs((df.v-df.v.mean())/df.v.std())>3)

M6. Create a DataFrame from a NumPy 2-D array. pd.DataFrame(np.arange(6).reshape(2,3), columns=list('abc'))

M7. Compute a correlation matrix. df.corr()

M8. Replace values conditionally with np.where. df['g']=np.where(df.v>0,'pos','neg')

M9. Bucket with np.select (multi-condition). df['t']=np.select([df.v<0, df.v==0],['neg','zero'],default='pos')

M10. Compute a weighted average. np.average(df.v, weights=df.w)

M11. Fill NaN with group median (transform). df['v']=df.groupby('g')['v'].transform(lambda s:s.fillna(s.median()))

M12. Vectorized distance from a target point. np.sqrt((df.x-0)**2 + (df.y-0)**2)

M13. Percentile rank of each row. df['v'].rank(pct=True)

M14. Convert wide numeric block to normalized rows. arr=df.to_numpy(); arr/arr.sum(1, keepdims=True)

M15. Random train/test split. msk=np.random.default_rng(0).random(len(df))<0.8; train,test=df[msk],df[~msk]