Prefill

Prefill is a javascript library to guess what an end-user wants to look up. The guess is based upon: current location and the time.

Why did I make this

Because a lot of journey planners exist, yet they all fail at delivering the simple user experience that a user would expect: prefill what a user will probably fill out.

Usage

General

We need the history in a certain format. Without further ado, this is an example of the only format we accept at this moment:

history = [
    {
        "datetime" : "2013-08-31T02:20Z", // iso8601
        "location" : { // the location of the user when it performed the action
            "longitude" : 3.14,
            "latitude" : 51.2
        },
        "from" : "http://stations.io/exampleidentifier1",
        "to" : "http://stations.io/exampleidentifier2"
    }//,
    //...
]

The history array will be used as the input for prefill:

p = new Prefill();
//this is a quite intense operation depending on how fast our neural network learns (oh yes, we're using a neural network)
p.prepare(history, function(){
    //this operation is okay
    result = p.guess("2013-08-31T02:20Z", 3.14, 51.2);
    console.log("I think " + result.to + " and " + result.from + " are the desired values");
});

Give it a try using node

In this repository, there is a test.js. If you have node.js installed and if you performed an npm install in this directory, you can run the test by launching:

./test.js

Use it in a browser

For the lazy or for those who just want to use prefill.js, you can find a ready made build in the build directory over here: https://github.com/OpenTransport/prefill/blob/master/build/bundle.js then,

• Step 1: load the script somewhere in : <script src="js/prefill.js"></script> or whatever you renamed to bundle.js file to.
• Step 2: Create an object somewhere: var p = new Prefill();

For those who want to develop further and use it in the browser, sorry, you will still need node.

• Step 0: clone this repository
• Step 1: npm install
• Step 2: (sudo) npm install -g browsify (if you don't have browserify installed yet)
• Step 3 option a: make (there's a makefile which atm will perform option b for you)
• Step 3 option b: browsify js/prefill.js -o build/bundle.js
• Step 4: use it in your HTML: <script src="build/bundle.js"></script>
• Step 5: Create an object somewhere: var p = new Prefill();

How it works

The history should be available on client side (e.g. in LocalStorage).

The history will be used to get all the necssary variables to train a neural network:

• Day of the week
• Day of the year
• Hours and minutes
• longitude and latitude of the users

and of course the "from" and "to" are included when training.

For the neural network, we will use brain.js.

Being smarter

In some cases we can be smarter than the neural network.

This happens when:

• the user's query history contains exactly 1 lookup in the near future (I suggest a timeframe of 2 hours). Because: a user may look up a certain route in the future and recheck the route when time approaches
• more?

Synchronizing the history

The history can be synchronized with a online user account. This is not necessary as everything can be stored in localstorage on client side.

License

MIT