insights 0.2.0

Quantitative algorithms, portfolio managers, data sources and contexts for Intuition

Insights
========

[](https://pypi.python.org/pypi/insights/)
[](https://travis-ci.org/hackliff/insights)
[](https://coveralls.io/r/hackliff/insights)
[](https://landscape.io/github/hackliff/insights/master)
[](https://requires.io/github/hackliff/insights/requirements/?branch=master)
[](https://pypi.python.org/pypi/insights/)

> Plug-and-play building blocks for modern quants

Quantitative algorithms, portfolio managers data sources, contexts and
middlewares for [Intuition][3]. Once provided to the configuration, those
modules, or the ones you create, are automatically used.

[Development Board][1]


Installation
------------

* One-liner

```console
$ wget -qO- http://bit.ly/1anxGhf | sudo FULL_INTUITION=true bash
```

* Step by step

```console
$ sudo apt-get install r-base
$ # Install R libraries
$ wget -qO- http://bit.ly/L39jeY | R --no-save
$ (sudo) pip install intuition

$ (sudo) pip install insights
```

Or if you plan to hack on it

```console
$ git clone https://github.com/hackliff/insights.git && cd insights
$ (sudo) pip install -r requirements.txt
$ # Install R code
$ cp -r R/ ~/.intuition
$ export PYTHONPATH=$PYTHONPATH:$PWD
```

Either way, you can use in your *intuition* configuration something like

```yaml
algorithm:
# Set here parameters you want to be accessible in the init function
notify: Nexus 5
long_window: 32
manager:
# And here those you will find under the $parameters dict in optimize()
max_weight: 0.3
modules:
manager: insights.managers.optimalfrontier.OptimalFrontier
algorithm: insights.algorithms.gradient.StochasticGradientDescent
data: insights.sources.live.EquitiesLiveSource
```

Check out readmes and source codes to learn what is available and the
parameters exposed. Documentation is on its way !


Code your own modules
---------------------

* [Algorithm API](https://github.com/hackliff/insights/blob/master/insights/algorithms/readme.md)
* [Portfolio API](https://github.com/hackliff/insights/blob/master/insights/managers/readme.md)
* [Data API](https://github.com/hackliff/insights/blob/master/insights/sources/readme.md)
* [Contexts](https://github.com/hackliff/insights/blob/master/insights/contexts/readme.md)
* [Middlewares](https://github.com/hackliff/insights/blob/master/insights/contexts/readme.md)

Here are some examples to get you started. To make your work available, just
feed the [Intuition][3] configuration (like above) a path it can find in the
environment variable `$PYTHONPATH`.

Everything is done to make writing modules easy, fun and efficient. So your
feedback, advices and contributions are happily welcome

* First, a classic buy and hold strategy, with a plugin which stores metrics in
[rethinkdb](www.rethinkdb.com):

```python
from intuition.zipline.algorithm import TradingFactory
import insights.plugins.database as database

class BuyAndHold(TradingFactory):
'''
Simpliest algorithm ever, just buy every stocks at the first frame
'''
def initialize(self, properties):

if properties.get('save'):
self.use(database.RethinkdbBackend(self.identity, reset=True)
.save_portfolio)

def event(self, data):
signals = {}

if self.day == 2:
for ticker in data:
signals[ticker] = data[ticker].price

return signals
```

* Here is the Fair manager example, which allocates the same weight in the
portfolio to all of your assets:

```python
from intuition.zipline.portfolio import PortfolioFactory

class Fair(PortfolioFactory):
'''
Dispatch equals weigths for buy signals and give up everything on sell ones
'''
def optimize(self, date, to_buy, to_sell, parameters):
# The algorithm fills 'to_buy' and 'to_sell' dicts by detecting
# buy and sell signals
# This dictionnary holds portfolio manager recommendations
allocations = {}

if to_buy:
# Split the portfolio in equal parts for each stock to buy
fraction = round(1.0 / float(len(to_buy)), 2)
for s in to_buy:
allocations[s] = fraction

for s in to_sell:
# Simply sells every stocks
allocations[s] = - self.portfolio.positions[s].amount

expected_return = 0
expected_risk = 1
return allocations, expected_return, expected_risk
```

* Now a source for backtests taking advantage of the awesome [Quandl][2] project.

```python
from intuition.zipline.data_source import DataFactory
from intuition.data.quandl import DataQuandl

class QuandlSource(DataFactory):
'''
Fetchs data from quandl.com
'''
def get_data(self):
return DataQuandl().fetch(self.sids,
start_date=self.start,
end_date=self.end,
returns='pandas')

@property
def mapping(self):
return {
'dt': (lambda x: x, 'dt'),
'sid': (lambda x: x, 'sid'),
'price': (float, 'Close'),
'volume': (int, 'Volume'),
'open': (int, 'Open'),
'low': (int, 'Low'),
'high': (int, 'High'),
}
```

[1]: https://trello.com/b/WvJDlynt/intuition
[2]: http://www.quandl.com/
[3]: https://github.com/hackliff/intuition