Updates

Stock Sentiment Analysis

We've currently only looked at numerical data as a trading signal, but there is a lot more information out in the world. A common source for signal is to use sentiment analysis

Sentiment Analysis uses natural language processing to attempt to detect sentiment in some text, for example:

  • This is bad! It is going to fail! (Negative sentiment)
  • Great job! It is a winner (Positive sentiment)

You can easily obtain positive/ negative sentiment score (-1.0 to +1.0) from a variety of sources on Quantopian! This sort of data is almost never free. Quantopian does have a free version of sentiment analysis (available only for years 2013-2015) provided by Accern Alphaone.

Accern Alphaone provides a sentiment from -1 to 1 and an impact score of 0 to 100. Impact is a measure of how likely a stock price will change as a result of the sentiment.

In the following lectures we will show you how to access this free data and set up a trading strategy with it.

Check out the video for full details.

Final Code

Here is the final code:

In [ ]:
# This section is only importable in the backtester
from quantopian.algorithm import attach_pipeline, pipeline_output

# General pipeline imports
from quantopian.pipeline import Pipeline
from quantopian.pipeline.factors import AverageDollarVolume


# Using the free sample in your pipeline algo
from quantopian.pipeline.data.accern import alphaone_free


def initialize(context):
    # Schedule our rebalance function to run at the start of each week.
    schedule_function(my_rebalance, date_rules.every_day())

    
    attach_pipeline(make_pipeline(), "pipeline")

def make_pipeline():

    
    # Screen out penny stocks and low liquidity securities.
    dollar_volume = AverageDollarVolume(window_length=20)
    
    # Grab the top 1000 most liquid stocks
    # is_liquid = dollar_volume.top(1000)
    is_liquid = dollar_volume.rank(ascending=False) < 1000
 
    # Add pipeline factors
    impact = alphaone_free.impact_score.latest
    sentiment = alphaone_free.article_sentiment.latest

    return Pipeline(columns={
            'impact': impact,
            'sentiment':sentiment,
            },
            screen = is_liquid)
            


    
def before_trading_start(context, data):
    port = pipeline_output('pipeline')
    
    # Grab stocks with 100 impact and >0.75 sentiment and go long.
    context.longs = port[(port['impact']==100) & (port['sentiment']>0.75)].index.tolist()
    
    # Grab stocks with 100 impact and <-0.75 sentiment and go short.
    context.shorts = port[(port['impact']==100) & (port['sentiment']< -0.75)].index.tolist()

    context.long_weight, context.short_weight = my_compute_weights(context)

def my_compute_weights(context):

    # Compute even target weights for our long positions and short positions.
    long_weight = 0.5 / len(context.longs)
    short_weight = -0.5 / len(context.shorts)

    return long_weight, short_weight

def my_rebalance(context, data):

    for security in context.portfolio.positions:
        if security not in context.longs and security not in context.shorts and data.can_trade(security):
            order_target_percent(security, 0)

    for security in context.longs:
        if data.can_trade(security):
            order_target_percent(security, context.long_weight)

    for security in context.shorts:
        if data.can_trade(security):
            order_target_percent(security, context.short_weight)