captn-client 2022.3.0rc1

Capt’n client

Capt’n python client 2022.3.0rc1

A python library encapsulating captn service REST API available at:

• https://api.airt.ai/

Docs

Full documentation can be found at the following link:

• https://docs.captn.ai/

How to install

If you don't have the captn library already installed, please install it using pip.

pip install captn-client

How to use

To access the captn service, you must create a developer account. Please fill out the signup form below to get one:

• https://bit.ly/3I4cNuv

Upon successful verification, you will receive the username/password for the developer account in an email.

Finally, you need an application token to access all the APIs in captn service. Please call the Client.get_token method with the username/password to get one.

You can either pass the username, password, and server address as parameters to the Client.get_token method or store the same in the CAPTN_SERVICE_USERNAME, CAPTN_SERVICE_PASSWORD, and CAPTN_SERVER_URL environment variables.

After successful authentication, the captn services will be available to access.

For more information, please check:

• Tutorial with more elaborate example, and

• API with reference documentation.

Below is a minimal example explaining how to load the data, train a model and make predictions using captn services.

!!! info

In the below example, the username, password, and server address are stored in **CAPTN_SERVICE_USERNAME**, **CAPTN_SERVICE_PASSWORD**, and **CAPTN_SERVER_URL** environment variables.

0. Get token

import json
from captn.client import Client, DataBlob, DataSource

Client.get_token()

1. Connect and preprocess data

In our example, we will be using the captn APIs to load and preprocess a sample CSV file stored in an AWS S3 bucket.

data_blob = DataBlob.from_s3(
    uri="s3://test-airt-service/sample_gaming_cohort_data/"
)
data_blob.progress_bar()

100%|██████████| 1/1 [01:35<00:00, 95.81s/it]

The sample data we used in this example doesn't have the header rows and their data types defined.

The following code creates the necessary headers along with their data types and reads only a subset of columns that are required for modeling:

prefix = ["revenue", "ad_revenue", "conversion", "retention"]
days = list(range(30)) + list(range(30, 361, 30))
dtype = {
    "date": "str",
    "game_name": "str",
    "platform": "str",
    "user_type": "str",
    "network": "str",
    "campaign": "str",
    "adgroup": "str",
    "installs": "int32",
    "spend": "float32",
}
dtype.update({f"{p}_{d}": "float32" for p in prefix for d in days})
names = list(dtype.keys())

kwargs = {"delimiter": "|", "names": names, "parse_dates": ["date"], "usecols": names[:42], "dtype": dtype}

Finally, the above variables are passed to the DataBlob.from_csv method which preprocesses the data and stores it in captn server.

data_source = data_blob.from_csv(
    index_column="game_name",
    sort_by="date",
    kwargs_json=json.dumps(kwargs)
)

data_source.progress_bar()

100%|██████████| 1/1 [00:40<00:00, 40.37s/it]

data_source.head()

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	date	platform	user_type	network	campaign	adgroup	installs	spend	revenue_0	revenue_1	...	revenue_23	revenue_24	revenue_25	revenue_26	revenue_27	revenue_28	revenue_29	revenue_30	revenue_60	revenue_90
0	2021-03-15	ios	Facebook Ads	Facebook Ads	campaign_144	adgroup_271	96	545.849976	130.000000	170.144928	...	632.002380	632.539246	660.857056	666.857056	666.857056	700.857056	729.035645	729.035645	1115.538574	1282.981079
1	2021-03-15	ios	snapchat_int	snapchat_int	campaign_569	adgroup_1634	10	207.919998	33.000000	39.000000	...	71.282143	71.282143	71.282143	71.282143	73.376190	73.376190	73.376190	73.376190	94.752373	143.727859
2	2021-03-15	ios	unityads_int	unityads_int	campaign_62	adgroup_275	41	73.339996	6.000000	22.297548	...	42.324429	42.462334	42.462334	42.462334	42.500793	43.640785	43.869404	43.871078	50.590389	50.590389
3	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_683	32	60.099998	33.000000	45.928074	...	58.466618	58.466618	58.466618	58.466618	58.466618	58.466618	58.466618	58.466618	58.466618	72.491997
4	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_278	9	19.500000	13.000000	13.438093	...	19.427818	19.427818	19.427818	19.427818	19.427818	19.427818	19.427818	19.427818	19.427818	19.427818
5	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_921	4	7.950000	0.000000	0.129259	...	4.205401	4.205401	4.205401	4.205401	4.205401	4.205401	4.205401	4.205401	4.227401	4.227401
6	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_327	3	5.380000	0.000000	0.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
7	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_5815	1	2.450000	0.000000	0.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
8	2021-03-15	ios	jetfuelit_int	jetfuelit_int	campaign_0	adgroup_162	1	0.600000	0.000000	0.026555	...	0.026555	0.026555	0.026555	0.026555	0.026555	0.026555	0.026555	0.026555	0.026555	0.026555
9	2021-03-15	android	Facebook Ads	Facebook Ads	campaign_23	adgroup_9	730	335.640015	36.300461	59.481831	...	124.508873	124.631615	124.644920	124.652664	124.905174	124.946205	125.000298	125.046783	125.483368	125.519073

10 rows × 41 columns

2. Training

# Todo