cuckookernel

A quick and dirty MATLAB to Julia translation/conversion reference guide.

This is not meant as a reference to the language. For that you should read the manual

Important Diferences

The first few are drawn from here

• Use brackets to index into vectors and matrices, i.e. do v[i] instead of v(i).
• Array assigment is done by reference, i.e after A = B modifying A will modify B!
• One dimensional vectors are column vectors by default.
• [17, 42, 25] and [17;42;25] both create a column vector. To create a row vector do [17 42 25], this is really a 1-by-3 (two-dimensional matrix).

cuckookernel

rfc = RandomForestClassifier( n_estimators=100, max_depth=5 )
rfc.fit( train_df, train_Y)
train_pred = rfc.predict( train_df )
accuracy = (train_pred == train_Y).sum() / len( train_pred )

cuckookernel

tc_exps, p0  = rf_explain( rfc, train_df)

cuckookernel

as_pyplot_figure( tc_exps[1], p0, train_df.columns, 'Passenger 1')

cuckookernel

import pandas as pd
from sklearn.ensemble import RandomForestClassifier

# tcxp.py and titanic_preproc.py are available at: 
#    https://github.com/YuxiGlobal/data-analytics/tree/master/tree_classif_explain
from tcxp import rf_explain, as_pyplot_figure
from titanic_preproc import preproc

train_df = pd.read_csv( "c:/tmp/titanic/train.csv" )
train_df, train_Y = preproc( train_df )

cuckookernel

checkins_by_loc = (checkins_df[['user_id', 'checkin_ts', 'location_id']] 
                      .set_index('location_id') ) 

chin_pairs = checkins_by_loc.join( checkins_by_loc, lsuffix='_ee', rsuffix='_er' ) 

pairs_filtered = (chin_pairs[(chin_pairs.checkin_ts_ee < chin_pairs.checkin_ts_er) &                                                 
                             (chin_pairs.user_id_ee != chin_pairs.user_id_er )]
                      .rename( columns= {"user_id_er" : "stalker",
                                         "user_id_ee" : "stalkee" }) 
                      .reset_index()

cuckookernel

import turicreate as tc

checkins = ( tc.SFrame.read_csv( 'Gowalla_totalCheckins.txt',                  
                                 delimiter='\t', header=False )
                .rename( {'X1': 'user_id', 'X2' : 'checkin_ts',
                          'X3': 'lat', 'X4' : 'lon',
                          'X5': 'location_id'} )
  [["user_id", "location_id", "checkin_ts"]] )

cuckookernel

chin_ps = ( checkins.join( checkins, on = 'location_id' )
                    .rename( {'checkin_ts'   : 'checkin_ts_ee',
                              'checkin_ts.1' : 'checkin_ts_er',
                              'user_id'      : 'stalkee' ,
                              'user_id.1'    : 'stalker' } ) )

pairs_filtered = chin_ps[ (chin_ps['checkin_ts_ee'] < chin_ps['checkin_ts_er']) &
                          (chin_ps['stalkee'] != chin_ps['stalker]) ]

cuckookernel

Inferred types from first 100 line(s) of file as
column_type_hints=[int,str,float,float,int]
------------------------------------------------------
Read 870755 lines. Lines per second: 520097
Finished parsing file /home/ubuntu/dask_experiment/Gowalla_totalCheckins.txt
Parsing completed. Parsed 6442892 lines in 5.43026 secs.

+---------+-----------+----------------+
| stalkee |  stalker  | location_count |
+---------+-----------+----------------+

cuckookernel

final_result = ( pairs_filtered[['stalkee', 'stalker', 'location_id']]
                    .unique()
                    .groupby( ['stalkee','stalker'] ,
                               {"location_count" : agg.COUNT })
                    .topk( 'location_count', k=5 )
                    .materialize() )

print( final_result )