CS109Site/index.html at master · dillontiner/CS109Site

History

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

<!DOCTYPE html>

<html>

<title>Uber Project</title>

$(function() {

$( "#datepicker" ).datepicker(

);

});

</script>

</head>

<body>

<span class="sr-only">Toggle navigation</span>

</button>

</div>

</div>

<li>

<option value="uberX">UberX</option>

<option value="uberXL">UberXL</option>

<option value="UberBlack">UberBlack</option>

<option value="UberSUV">UberSUV</option>

<option value="uberTAXI">UberTAXI</option>

<option value="uberSELECT">UberSELECT</option>

</select>

</form>

</li>

<li>

</form>

</li>

<li>

</form>

</li>

<li>

<option value="Atlanta">Atlanta</option>

<option value="Boston">Boston</option>

<option value="San Francisco">San Francisco</option>

<option value="Washington DC">Washington DC</option>

</select>

</form>

</li>

<p id="submitChangesText"> What's my surge? </p>

</button>

</li>

</ul>

</div>

</div>

</nav>

<h1><b id="selectedCity">Select a city</b></h1>

</div>

</div>

</div>

</div>

</div>

</div>

<h2>No Surge</h2>

</div>

<h2>Low Surge</h2>

</div>

<h2>Mid Surge</h2>

</div>

<h2>High Surge</h2>

</div>

</div>

<!--div class="row row-border">

</div>

</div>

</div-->

<h1><b>Surges Explained</b></h1>

</div>

<h3><b>No surge:</b> Looks like it's your lucky day!</h3>

<h3><b>Mid surge:</b> 1.5x</h3>

<h3><b>Low surge:</b> Under 1.5x</h3>

<h3><b>High surge:</b> Over 1.5x</h3>

</div>

</div>

</div>

<p>First, we collected the data using uber's new API on an AWS machine. Second, we used exploratory data analysis to glean more insights into the data set. Much of the exploratory data analysis proved our intuition correct. For example, the colder the temperature, the less likely people are to walk and the more likely there will be a surge price on ubers. Finally, we trained a Random Forest Classifier on our dataset to have a comprehensive model that can be used to glean further insights into the data and create a very reliable prediction of the surge price (over 96% accuracy!). After creating the model, we wanted to look at Boston specifically. We turned attention to two of the most important features we found in our model: temperature and car type. We proved our intuition with temperature using the model, then showed that it is possible also show consumer behaivor in Boston with respect to service type.</p><br>

</div>

<h1><b>What got us going</b></h1>

</div>

<p>We wanted to explore Uber’s concept of surge pricing by conducting an analysis of which factors affect surge pricing the most. This was particularly interesting to us because Uber’s API was just recently made public, so there have not been many predictive/analytical projects looking at surge pricing. This gives us a novel opportunity to explore data that affects our daily lives, and to incorporate several of the statistics and computational methodologies that we learned this semester. The data set that we have compiled can also allow us to look at many different variables and their effect on surge pricing. For example, we will look at how temperature and weather affect the surge charge on an uber. Also, should someone call the Uber now, or if they have time wait an hour or two in order to try and find a better price? We all use Uber, and love the service; however, we wanted to create a predictive model in order to find the times that we do not have to be charged extra money!</p><br>

</div>

</div>

</div>

<p>The Uber API was made available to the general public earlier this year. While the company's main motivation for opening up the API was to allow technology developers to embed Uber into their apps, it also allowed for users like to amass and analyze ride data. While Uber did not allow us to access past historical data, once we knew we wanted to collect data we were able to use AWS to make calls (queries?) every 5 minutes to the Uber API and receive back information as if we were a user requesting a ride at that moment in time. We did this over 3 weeks in 5 different cities.</p><br>

</div>

</div>

<p> The dataset we collected over 3 weeks was information pulled from an API call every 5 minutes to Uber's <a href="https://developer.uber.com/docs/v1-estimates-price" target="blank">Price Estimates</a> endpoint passing in parameters for the location we wanted to request a ride from, as well as an API call to WeatherUnderground's API to receive temperature and weather data. For each call we received the unique ride ID, time/date, type of uber, surge, temperature, weather description, state, and income of the area. </p><br>

</div>

<h1><b>What we found</b></h1>

</div>

<h3>Exploratory Analysis</h3>

<p>We found that generally most rides do not have surge, only about 4% of the time will an Uber ride contain a surge charge. Average surge overall was about 1.024x the base price. The graphs we found to be most interesting were the graph on the effect of the city effect on surge, the effect of temperature on surge, the effect of time of day on surge, and the effect of weather type on surge.</p>

<p> Of all the cities that we looked at Boston and Washington DC have highest overall surge and New York has the lowest. New York having the lowest surge intuitively makes sense as New York is an Uber hub as well as has lots of alternative modes of transportation, which limits excess demand. Boston and DC have less developed public transportation systems so the demand for Uber is higher. </p>

<br>

<h3>Temperature</h3>

<p> We see that with extremely cold temperatures there is a higher surge because of higher demand for Uber with not as much supply. We see a similar effect but not as strong for extremely hot temperatures. </p>

<br>

<h3>Weather</h3>

<p> Weather conditions are clear most of the time, but when there is rain or there is an indication of rain with cloudiness, surge pricing goes up as people do not want to travel in adverse weather conditions. </p>

<br>

<p> Overall, the highest average surge is at 5 and 6 PM, followed by 8 AM. Since these are rush hour times, it makes sense that there would be high demand for Ubers at these times, and not enough supply. The lowest average surge occurs at 10 AM and 8 PM, which also makes sense because there is generally a low demand for transportation at these times of the day. </p>

<br>

<h3>Model Summary</h3>

<p>In order to create a predictive model, we decided to use a Random Forest Classifier. Using a Random Forest Classifier makes sense for three reasons. First, we are using a dataset with unbalenced classes in that there are many more cases where there is "No Surge" than the other three classes. A Random Forest works well with unbalenced classes. Second, we are using a lot of data in the training of the model. Random Forests samples from the data many times and so the prediction will be optimal if there is a lot of data to be used. Finally, we are able to look at feature importance in the data, which is particularly enlightening on this dataset. Thus we are able to see what features have the biggest effect on the pricing prediction.</p>

<p>We split the dataset into a test and training set first in order to test for overfitting of the model. We then had to choose the hyperparameters for the model. We used a 5 fold cross validation technique in order to train the hyperparameters of the model. After fully training the model, we found that the model predicted the correct class 96.3% of the time on the test set and 98% of the time on the training set. This small difference shows that the model was not overfitted and we properly trained the model.</p>

<br>

<h3>Final Analysis</h3>

<p>For the final Analysis of the model, we decided to look at two aspects of the model. First, we look at the feature importance found in the model sorted by the features that affect the prediction the model the most. Second, we restrict the model to the Boston area to look at how the city specific surge prices are affected by some of the most important variables found in the model: temperature and cartype</p>

<br>

<h3>Feature Importance</h3>

<p>The graph shows the top 10 most predictive features in the model. As you can see, temperature is by far the most predictive variable that we pass into the model. Second, the next variable that seems to affect the prediction is the type of car. This is interesting because our car type feature serves as a proxy for the supply and demand dynamics of each uber service.</p>

<br>

<h3>Temperature's Effect on Boston Ubers</h3>

<p>We created dummy data in order to create this plot. We fixed all other variables but only altered the temperature variable to see how the probability of a surge changes. As you can see, as temperature increases so does the probability of there being no uber surge. Interestingly enough, if you want to get a cheap uber, then use the service when it is hot outside! The model predicts that the colder the temperature is, the more likely there is to be a surge multiple.</p>

<br>

<h3>Car Type Effect on Boston Ubers</h3>

<p>The graph shows an interesting proxy into consumer usage of uber's services in Boston. We interpret the graph as the higher the probability of a surge, the more demand there is for that particular service, which in turn leads to higher surges. uberX and uberXL are by far the most demanded services in the Boston area. uberBLACK seems to be the least demanded service. The graph also shows the widespread application of the model and the incredible amount of analysis that can be done to glean insights into uber's business!</p>

<br>

</div>

<h1><b>Meet the drivers</b></h1>

</div>

</div>

<b>

<h4 class="riderName">Peter Traber</h4>

Applied Math & Government

</b>

</div>

</div>

<b>

<h4 class="riderName">Abbie Orlando</h4>

Neurobiology & Computer Science

</b>

</div>

</div>

<b>

<h4 class="riderName">Summer Carter</h4>

Computer Science & Economics

</b>

</div>

</div>

<b>

<h4 class="riderName">Dillon Tiner</h4>

Statistics & History of Art and Architecture

</b>

</div>

</div>

<b>

<h4 class="riderName">Scratchy</h4>

Astrophysics & Folklore and Mythology (Joint)

</b>

</div>

/* Location Finder -- NECESSARY?? */

var x = document.getElementById("suggestCityText");

function getLocation() {

if (navigator.geolocation) {

navigator.geolocation.getCurrentPosition(showPosition, showError);

} else {

x.innerHTML = "Geolocation is not supported by this browser.";

}

function showPosition(position) {

x.innerHTML = "Thanks!";

}

function showError(error) {

switch(error.code) {

case error.PERMISSION_DENIED:

x.innerHTML = "Alrighty then!"

break;

case error.POSITION_UNAVAILABLE:

x.innerHTML = "Thanks!"

break;

case error.TIMEOUT:

x.innerHTML = "Thanks!"

break;

case error.UNKNOWN_ERROR:

x.innerHTML = "Thanks!"

break;

}

</script>

</body>

</footer>

</html>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

index.html

index.html

Files

index.html

Latest commit

History

index.html

File metadata and controls