Running Python in Hive/Hadoop

Standard

One of the things I love about running Hive is the ability to run Python and leverage the power of the parallel processing. Below I’m going to show a stripped down example of how to integrate a Hive statement & Python together to aggregate data to prepare it for modeling. Keep in mind, you can also use Hive & Python to transform data line by line as well, and it extremely handy for data transformation.

Use case: print out an array of products sold to a particular user. Again is a basic example, but you can build upon this and generate products sold for every user, then use KNN to generate clusters of users, or perhaps Association Rules to generate baskets.

Here is the Python script, which will have to be saved in local Hadoop path:


#!/usr/bin/python
import sys

items_sold = []  # create global list variable

class Items:  # create class to store and access items added
    def __init__(self, x):
    	self.x = x

    def set_x(self, x):
        self.x = x
    
    def get_x(self):
        return self.x

def print_results():  # print output in Hive
	result_set = [item.get_x() for item in items_sold];
	print (result_set)

	# Hive submits each record to stdin
	# The record/line is stripped of extra characters and submitted
for line in sys.stdin:
	line = line.strip()
	purchased_item = line.split('\t')
	items_sold.append(Items(purchased_item))

print_results()

Here is the hive statement:


add file blog_hive.py; 
select TRANSFORM (a.purchased_item)
using 'blog_hive.py'
AS array_purchased
from (select purchased_item from company_purchases where user_id = 'u1') a;

Result in Hive will be similar to this: [‘s_123’, ‘s_234’, ‘s890’]

Probability of a Revenue Threshold

Standard

A retailer’s website purchases have an average order size of $100 and a standard deviation of $75. What is the probability of 10 orders generating over $1,250 in Revenue?

mean = $100.00
stdev = $75.00

avg_order_needed = $1250/10 = $125.00
standard_error = $75/sqrt(10) = $23.72
z-score = (125.00 – 100.00)/23.72 = 1.05

We are looking to solve for this shaded area under the curve.

upper_tail_post

Looking up on z-table for 1.05, the probability is 0.1469 or 14.7% of a obtaining $1,250 in Revenue from 10 random orders.

Crowd-sourced Recommender Demo

Standard

Recommender Demo – click here!

This demo of a recommender is to illustrate an example of how a website (online music, e-commerce, news) generates recommendations to increase engagement and conversions.

This is not production ready, merely a POC of how it works.

* user selects favorite activities
* data is passed to server and processed in hadoop
* user can go to results page and select an activity to get recommendations

At this point, an auto-workflow has not been built, so there are a series of steps to create the new dataset. Here are the general steps:

1. user data feeds into database via website (which is used in generating recommendations)
2. data is moved and process in Hadoop
3. data is moved to MySQL, accessible using PHP
4. user selects an activity, and the crowd-sourced recommendations are displayed

Example: How Crowd-Sourcing Works (co-occurrence recommendations) Using Activities

All Users Activity History
| Activity | Art Fair | Fishing | Shovel Snow | Wedding |
| Jon          | Yes           | Yes         | Yes                      | No              |
| Jane        | No            | Yes         | No                        | Yes            |
| Jill            | Yes           | Yes         | No                        | Yes            |

A New User like to go to Weddings, and we need to recommend them other activities:
* Find Wedding in History Matrix who also enjoyed Wedding to it: U{Jane, Jill}
* Identify other activities same users (U) enjoyed, and rank by count

Recommendation
| Activity | Rank | Count of User (co-occurrence |
| Fishing  |  1         |  2                                                               |
| Art Fair |  2         | 1                                                                |

Predictive Algorithms on Million Song Dataset

Standard

I’ve had the opportunity within a Data Mining course in my graduate Software Engineering program to be part of a project in which we were to create a “recommendation engine”. The dataset we used was called the which there are 1M songs, along with play history of 380k users.

The goal was to provide a recommendation (ranked 1-10) of songs based on a current song played. We used three algorithms, Association Rules, Naive Bayes, and user-user co-occurance. When tested, the results were mixed, with Association Rules providing the top F1 scores, but also had the lowest # of recommendations (for a large portion of songs had less than 10 songs recommended). Co-occurance was close behind with the 2nd best F1 score, and provided the largest output of songs, as well as the lowest requirement of computational requirements.

Here is the full project on github.