Clinical Analytics API

Getting Started

API: https://clinical-analytics-api.herokuapp.com/ca/getpredictions

Scope

Scope Create and host an API that will serve the results from a prebuilt Python or R model (either one API for the Python model or one for the R model). You can decide on your own JSON format for the requests. The models can score multiple records in a single call and the API should do so as well. We have provided a CSV file for both languages with a sample of the inputs and outputs required. In both cases, the ‘Predictions’ column is the output that should be returned by the API

Technical Specifications

_	_
Operating System	Ubuntu 18.04
Development Language(s)	Python==3.6.5
Cloud Platform	Heroku/Heroku-18 stack
Packages/Libraries	numpy==1.15.4
	pandas==1.1.0
	Flask==1.1.0
	scipy==1.5.4
	scikit-learn==0.19.1
	gunicorn==20.1.0

Data I/O Format Design

The data input and outut of the API

API Input JSON

The API expects a JSON format input of the following example structure

{
    "Pclass":[2, 3], 
    "Sex": [1, 1], 
    "Age":[24, 61], 
    "SibSp":[0, 0], 
    "Parch":[0, 0],
    "Fare":[13, 6.2375], 
    "Embarked_S":[1, 1], 
    "Embarked_C":[1, 0]
}

The inputs need to be followed the list of the same specified datataypes as mentioned in the above features table not scalars. Check out more acceptable Json input formats in HOW-to-Guide

API Output JSON

{
    "prediction": [
        0.38441490346019563,
        0.17907895626397127,
        0.4239311987959022,
        0.39193345612918523,
        0.3977242630764239
    ]
}

Solution Approach

Develop the ml api using python flask framework
Containerise the application using docker
Deploy the container into heroku cloud
Test the final deployment using post man and requests

Cloud Deployment

The deployment methods available for Heroku are

Heroku Git (Heroku CLI )
Github
Container Registry ( Heroku CLI ) Base Docker  Used Method

From the above methods the use of Github is not recommended as the current version needs to be deployed on Heroku-18 stack (Ubuntu-18.04) where by default heroku builds on Heroku-20 stack

The suited mode of deployments are Heroku Git and Container Registry using the Heroku CLI

The Container method is the ideal method for the deployment for future proofing the deployment pipeline and ease of development. Deployment Steps

# Commands in command line 

$ heroku login
$ heroku container:login
$ heroku container:push web -a clinical-analytics-api
$ heroku container:release web -a clinical-analytics-api

Tests

Testing with POSTMAN

The API calls can be tested using the POSTMAN

Step 1 : goto the Postman API Platform
Step 2 : navigate to the create new section by clicking on “Create new”
Step 3: in the new page select POST and add https://clinical-analytics-api.herokuapp.com/ca/getpredictions In the address bar
Step 4: select the Body tab and datatype JSON paste the input test JSON in the below text area
Step5: Click send
Step 6: in the response if 200 Is found it means its success and in the results body select JSON in the dropdown . The predictions output is found like below image

Python Requests

The API method can be requested using the python requests package

How-To-Guide

Get predictions for a single record of values

import requests

url = "https://clinical-analytics-api.herokuapp.com/ca/getpredictions"
Input_json = {
    "Pclass": [3], 
    "Sex":  [1], 
    "Age":[24], 
    "SibSp":[0], 
    "Parch":[0], 
    "Fare":[6.2375], 
    "Embarked_S": [1], 
    "Embarked_C":[1]
}

r = requests.post(url, json = Input_json)
r.text.strip()
print(r)
print(r.json())

Output:

{'prediction': [0.33116930987918997]}

Get predictions for a multiple records of values

import requests

url = "https://clinical-analytics-api.herokuapp.com/ca/getpredictions"

Input_json = {
 "Pclass":{"0":2,"1":3,"2":2,"3":2,"4":3},
 "Sex":{"0":False,"1":True,"2":False,"3":True,"4":False},
 "Age":{"0":24.0,"1":61.0,"2":17.0,"3":18.0,"4":24.0},
 "SibSp":{"0":0,"1":0,"2":0,"3":0,"4":0},
 "Parch":{"0":0,"1":0,"2":0,"3":0,"4":3},
 "Fare":{"0":13.0,"1":6.2375,"2":12.0,"3":11.5,"4":19.2583},
 "Embarked_S":{"0":True,"1":True,"2":False,"3":True,"4":False},
 "Embarked_C":{"0":False,"1":False,"2":True,"3":False,"4":True}
}

r = requests.post(url, json = json_d)
r.text.strip()
print(r)
print("\n")
print(r.json())

Output:

<Response [200]>


{'prediction': [0.38441490346019563, 0.17907895626397127, 0.4239311987959022, 0.39193345612918523, 0.3977242630764239]}

Get predictions for a DataFrame

import pandas as pd
import requests
import json

df = pd.read_csv('data/test_data_for_candidate_python.csv')
__input = df[['Pclass', 'Sex', 'Age', 'SibSp', 'Parch', 'Fare', 'Embarked_S', 'Embarked_C']]
Input_json = __input.to_dict()
url = "https://clinical-analytics-api.herokuapp.com/ca/getpredictions"
r = requests.post(url, json = Input_json)
r.text.strip()
print(r)
print("\n")
print(r.json())

Output:

{'prediction': [0.38441490346019563, 0.17907895626397127, 0.4239311987959022, 0.39193345612918523, 0.3977242630764239, 0.3453320907856661, 0.7171907356545539, 0.36509353094892455, ----------------------------------------------------0.46576400760705566, 0.3548613897216256, 0.3831127920903322, 0.4796898111659956, 0.33217626419946283, 0.3810110273939012, 0.3415961034276208]}

kasichennupati / clinical-analytics-api Goto Github PK