Skip to content

Home

Jump to bottom
lubos31262 edited this page May 21, 2021 · 37 revisions

Feature collector "price-fc"

Version: 1.0

Date: 14.04.2021

Authors: Ľuboš Buzna; Milan Straka

Address: University of Žilina, Univerzitná 8215/1, 010 26 Žilina, Slovakia

Description

The "price-fc" feature collector is a module of the Ride2Rail Offer Categorizer responsible for the computation of the following determinant factors: "total_price", "ticket_coverage" and "can_share_cost".

Computation can be executed from "price.py" by running the procedure extract() which is binded under the name compute with URL using FLASK (see example request below). Computation is composed of four phases:

Phase I: Extraction of data required by price-fc feature collector from the cache. A dedicated procedure defined for this purpose from the unit "cache_operations.py" is utilized.

Phase II: Use of the external service to convert prices to EUR (if needed). A class collecting external data has been implemented in the module "exchange_rates.py".

Phase III: Computation of weights assigned to "price-fc" feature collector. For the aggregation of data at the tripleg level and for normalization of weights a dedicated procedure implemented in the unit "normalization.py" are utilized. By default "z-scores" are used to normalize data.

Phase IV: Storing of the results produced by "price-fc" to cache. A dedicated procedure defined for this purpose in the unit "cache_operations.py" is utilized.

Configuration

The following values of parameters can be defined in the configuration file "price.conf".

Section "running":

  • "verbose" - if value "1" is used, then feature collector is run in the verbose mode,
  • "scores" - if value "minmax_score" is used, the minmax approach is used for normalization of weights, otherwise, the "z-score" approach is used.

Section "cache":

  • "host" - host address where the cache service that should be accessed by "price-fc" feature collector is available
  • "port" - port number where the cache service that should be accessed used by "price-fc" feature collector is available

Example of the configuration file "price.conf":

[service]
name = price
type = feature collector
developed_by = Lubos Buzna <https://github.com/lubos31262> and Milan Straka <https://github.com/bioticek>

[running]
verbose = 1
scores  = minmax_scores

[cache]
host = cache
port = 6379

Usage

Local development (debug on)

The feature collector "price-fc" can be launched from the terminal locally by running the script "price.py":

$ python3 price.py
 * Serving Flask app "price" (lazy loading)
 * Environment: development
 * Debug mode: on
 * Running on http://127.0.0.1:5001/ (Press CTRL+C to quit)
 * Restarting with stat
 * Debugger is active!
 * Debugger PIN: 441-842-797

Moreover, the repository contains also configuration files required to launch the feature collector in Docker from the terminal by the command docker-compose up:

docker-compose up
Starting price_fc ... done
Attaching to price_fc
price_fc    |  * Serving Flask app "price.py" (lazy loading)
price_fc    |  * Environment: development
price_fc    |  * Debug mode: on
price_fc    |  * Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)
price_fc    |  * Restarting with stat
price_fc    |  * Debugger is active!
price_fc    |  * Debugger PIN: 250-384-212

Example Request

To make a request (i.e. to calculate values of determinant factors assigned to the "price-fc" feature collector for a given mobility request defined by a request_id) the command curl can be used:

$ curl --header 'Content-Type: application/json' \
       --request POST  \
       --data '{"request_id": "123x" }' \
         http://localhost:5001/compute

API for exchange rates:

Implemented API is reading exchange rates from the service provided by the European Central Bank https://www.ecb.europa.eu/stats/eurofxref/eurofxref-daily.xml

Clone this wiki locally