1 00:00:00,000 --> 00:00:04,000 So now let's go ahead and probably implement this regression problem statement. 2 00:00:04,000 --> 00:00:06,000 So what I will do I will go ahead and write over here. 3 00:00:06,000 --> 00:00:12,000 And inside this folder only I will be creating this because we are going to follow the same steps. 4 00:00:12,000 --> 00:00:20,000 So I'll go ahead and write I regression regression ipynb I'll say churn regression or I'll say salary 5 00:00:20,000 --> 00:00:21,000 regression. 6 00:00:21,000 --> 00:00:21,000 Okay. 7 00:00:21,000 --> 00:00:23,000 So this will be my file. 8 00:00:23,000 --> 00:00:27,000 I will go ahead and quickly select my kernel okay. 9 00:00:27,000 --> 00:00:30,000 Now first of all, what are the steps that we will do. 10 00:00:30,000 --> 00:00:36,000 So initially we will go ahead and import all the necessary modules that is required like train test, 11 00:00:36,000 --> 00:00:39,000 split, standard scaler, label encoder one hot encoder and pickle. 12 00:00:39,000 --> 00:00:40,000 Okay. 13 00:00:40,000 --> 00:00:41,000 So we'll go with the first step over here. 14 00:00:42,000 --> 00:00:47,000 Then uh, after this we will be reading our CSV file. 15 00:00:47,000 --> 00:00:49,000 So I'll go ahead and write read underscore CSV. 16 00:00:49,000 --> 00:00:54,000 And let me just go ahead and write this as churn modeling. 17 00:00:54,000 --> 00:00:55,000 Churn. 18 00:00:56,000 --> 00:01:00,000 Churn underscore modeling dot csv file okay. 19 00:01:01,000 --> 00:01:07,000 Now if I go ahead and write data dot head I will be able to get my out input over here. 20 00:01:07,000 --> 00:01:09,000 And this is my entire data set right. 21 00:01:09,000 --> 00:01:15,000 As I said estimated salary will be my output feature and remaining all will be my independent feature. 22 00:01:15,000 --> 00:01:15,000 Okay. 23 00:01:15,000 --> 00:01:18,000 And that is the reason the output feature is a continuous value over here. 24 00:01:18,000 --> 00:01:22,000 So we are considering this as a regression problem statement. 25 00:01:22,000 --> 00:01:29,000 Now after this the next state that what I will do is that I will delete this row number, customer ID 26 00:01:29,000 --> 00:01:30,000 and surname. 27 00:01:30,000 --> 00:01:36,000 Whatever things we did in the classification problem almost the same way, we have to probably do this. 28 00:01:36,000 --> 00:01:37,000 Okay, so I have executed this. 29 00:01:37,000 --> 00:01:45,000 Then we will go ahead and convert this, uh, gender from female male to zeros and ones. 30 00:01:45,000 --> 00:01:48,000 Because this here I'm going to use my label encoder. 31 00:01:48,000 --> 00:01:52,000 So for that I will go ahead and create my variable label encoder gender. 32 00:01:52,000 --> 00:01:54,000 I will initialize to label encoder. 33 00:01:54,000 --> 00:01:55,000 And then I will go ahead and write. 34 00:01:55,000 --> 00:01:59,000 Data of gender is equal to label encoder dot fit underscore transform okay. 35 00:01:59,000 --> 00:02:03,000 Similarly I will go ahead and do it for the geographic column. 36 00:02:03,000 --> 00:02:08,000 Also where I'm going to use one hot encoder Again I'll use fit underscore transform. 37 00:02:08,000 --> 00:02:10,000 I'll convert this into arrays. 38 00:02:10,000 --> 00:02:14,000 I will get all my values, then convert this into my data frame along with my features. 39 00:02:14,000 --> 00:02:20,000 Okay, so if you go ahead and see your geo underscore encoded underscore DF, you'll be able to see 40 00:02:20,000 --> 00:02:22,000 hey I'm getting this three features. 41 00:02:22,000 --> 00:02:23,000 Okay. 42 00:02:23,000 --> 00:02:28,000 Then we are going to combine this one hot encoded columns with my original data. 43 00:02:28,000 --> 00:02:30,000 Step by step we are going to do this. 44 00:02:30,000 --> 00:02:33,000 So finally you'll be able to see data dot head. 45 00:02:33,000 --> 00:02:35,000 And we also dropping the geographic column. 46 00:02:35,000 --> 00:02:37,000 Now I don't have my geographic column. 47 00:02:37,000 --> 00:02:41,000 And if you probably go ahead and see my this three columns have got added. 48 00:02:41,000 --> 00:02:41,000 Perfect. 49 00:02:42,000 --> 00:02:48,000 Now we are going to split our data into features uh into our independent features and target feature. 50 00:02:48,000 --> 00:02:49,000 Right dependent feature also. 51 00:02:49,000 --> 00:02:52,000 So I'm dropping estimated salary from this. 52 00:02:52,000 --> 00:02:57,000 And all the remaining features will go in x axis and my estimated salary will go in the y axis. 53 00:02:57,000 --> 00:02:57,000 Okay. 54 00:02:58,000 --> 00:03:00,000 So this is done a bit of processing. 55 00:03:00,000 --> 00:03:06,000 Now I will save all the pickle file specifically for transformation like label encoder dot pickle. 56 00:03:06,000 --> 00:03:08,000 Then you have one hot encoder dot pickle. 57 00:03:08,000 --> 00:03:12,000 Then one more pickle that we are going to probably create a scalar pickle. 58 00:03:12,000 --> 00:03:20,000 But uh, the scalar dot pickle we will specifically use uh only when we try to, uh, you know, do 59 00:03:20,000 --> 00:03:21,000 the train test split. 60 00:03:21,000 --> 00:03:23,000 So let's go ahead and do the train test split also. 61 00:03:23,000 --> 00:03:26,000 So here I will take this code quickly. 62 00:03:26,000 --> 00:03:31,000 I will just copy and paste it so that I don't have to waste more time on this writing the entire thing, 63 00:03:31,000 --> 00:03:32,000 because we are doing the same thing. 64 00:03:32,000 --> 00:03:37,000 If I was doing something else then I could had probably done that. 65 00:03:37,000 --> 00:03:41,000 Okay, so here I'll go back to my salary regression quickly. 66 00:03:41,000 --> 00:03:44,000 I will go ahead and write code now. 67 00:03:44,000 --> 00:03:45,000 Quickly I'll paste it over here. 68 00:03:45,000 --> 00:03:45,000 Okay. 69 00:03:45,000 --> 00:03:46,000 Train test split. 70 00:03:46,000 --> 00:03:54,000 And now I will go ahead and apply my standard scaler scale this feature quickly. 71 00:03:55,000 --> 00:03:56,000 Salary okay. 72 00:03:56,000 --> 00:03:59,000 And here I will go ahead and do this. 73 00:03:59,000 --> 00:04:01,000 And this is done right. 74 00:04:01,000 --> 00:04:03,000 Okay I'm getting an error. 75 00:04:03,000 --> 00:04:04,000 Could not convert string to me. 76 00:04:04,000 --> 00:04:05,000 Why. 77 00:04:06,000 --> 00:04:12,000 Because let's see my x x x value. 78 00:04:12,000 --> 00:04:12,000 Okay. 79 00:04:12,000 --> 00:04:14,000 Gender is still male and female. 80 00:04:14,000 --> 00:04:15,000 Why why why why. 81 00:04:15,000 --> 00:04:19,000 Because data of gender it is. 82 00:04:19,000 --> 00:04:19,000 Okay, fine. 83 00:04:20,000 --> 00:04:22,000 Data of gender. 84 00:04:23,000 --> 00:04:28,000 So if I replace this what is my data over here. 85 00:04:30,000 --> 00:04:35,000 If you see the data or let me execute this once again, I think I have missed something so I'll quickly 86 00:04:35,000 --> 00:04:36,000 execute this. 87 00:04:36,000 --> 00:04:37,000 I'll remove this tree. 88 00:04:37,000 --> 00:04:38,000 I'll see this. 89 00:04:38,000 --> 00:04:40,000 Now let's go ahead and check my data. 90 00:04:40,000 --> 00:04:43,000 So my data is basically having zeros and ones. 91 00:04:43,000 --> 00:04:47,000 I think I missed one line of code in executing it. 92 00:04:47,000 --> 00:04:48,000 Then I'll do the concatenation. 93 00:04:48,000 --> 00:04:51,000 And finally I will do the concatenation. 94 00:04:51,000 --> 00:04:55,000 This has got executed I have gender zero and 1X1Y. 95 00:04:55,000 --> 00:05:00,000 I get it and I'm doing the train test split and now it should do the scaling. 96 00:05:00,000 --> 00:05:01,000 Okay, so I missed one code. 97 00:05:01,000 --> 00:05:03,000 So that is the reason that was the issue. 98 00:05:03,000 --> 00:05:05,000 Now we'll go ahead and save all the pickle files. 99 00:05:05,000 --> 00:05:10,000 Scaler dot pickle file one hot encoder G0 and label encoder gender pickle file. 100 00:05:10,000 --> 00:05:19,000 Okay, now it's the time to train our Ann with regression problem statement. 101 00:05:19,000 --> 00:05:19,000 Right. 102 00:05:21,000 --> 00:05:24,000 How to train our Ann over here. 103 00:05:24,000 --> 00:05:26,000 That is what we are basically going to discuss. 104 00:05:26,000 --> 00:05:27,000 And this is the important thing. 105 00:05:27,000 --> 00:05:29,000 Remaining thing are almost same. 106 00:05:29,000 --> 00:05:31,000 We do not have to make any changes okay. 107 00:05:32,000 --> 00:05:34,000 Now let's go ahead and quickly do this. 108 00:05:34,000 --> 00:05:39,000 So first of all again I will import TensorFlow as Keras okay. 109 00:05:39,000 --> 00:05:41,000 And then we are importing sequential and dense. 110 00:05:41,000 --> 00:05:43,000 So first of all we'll go ahead and import this. 111 00:05:43,000 --> 00:05:48,000 So once this is imported we are good to go and work with sequential and dense. 112 00:05:48,000 --> 00:05:52,000 And we are good to work with our entire neural network okay. 113 00:05:52,000 --> 00:05:54,000 So here we are executing this. 114 00:05:54,000 --> 00:05:58,000 Then quickly let us go ahead and do this okay. 115 00:05:58,000 --> 00:06:00,000 So we will go ahead and build our model again. 116 00:06:00,000 --> 00:06:03,000 To build our model we will use the same technique. 117 00:06:04,000 --> 00:06:09,000 Like first of all I will say hey go ahead and take this extreme dot shape of one as input shape. 118 00:06:09,000 --> 00:06:11,000 This will be my input layer. 119 00:06:11,000 --> 00:06:13,000 Then we are going to apply an activation function. 120 00:06:13,000 --> 00:06:15,000 The first hidden layer will have 64 neurons. 121 00:06:15,000 --> 00:06:17,000 The second hidden layer will be having 32 neurons. 122 00:06:17,000 --> 00:06:19,000 And the activation function will be ReLU. 123 00:06:19,000 --> 00:06:21,000 And the dense will basically have one. 124 00:06:21,000 --> 00:06:22,000 Okay. 125 00:06:22,000 --> 00:06:26,000 Now when we are using one here, we are not applying any activation function. 126 00:06:26,000 --> 00:06:30,000 When we don't apply any activation function, the default activation function that will be applied is 127 00:06:30,000 --> 00:06:32,000 called as linear activation function. 128 00:06:32,000 --> 00:06:35,000 Now linear activation function is specifically for regression. 129 00:06:35,000 --> 00:06:36,000 That basically means I'm not doing anything. 130 00:06:36,000 --> 00:06:38,000 Instead I'm getting whatever output I'm getting. 131 00:06:38,000 --> 00:06:39,000 I'm using that okay. 132 00:06:40,000 --> 00:06:44,000 Now after this we will go ahead and compile this model okay. 133 00:06:44,000 --> 00:06:46,000 So we will go ahead and write compile the model. 134 00:06:46,000 --> 00:06:50,000 Now for compiling the model I will go ahead and write compile model dot compile. 135 00:06:50,000 --> 00:06:54,000 And here uh first of all I'll go ahead and write my optimizer. 136 00:06:54,000 --> 00:06:56,000 Let's say the optimizer is Adam. 137 00:06:56,000 --> 00:07:00,000 For this case you can also go ahead and initialize it from Tensorflow.keras. 138 00:07:00,000 --> 00:07:01,000 But I'm directly writing the keyword. 139 00:07:02,000 --> 00:07:04,000 Then my loss. 140 00:07:04,000 --> 00:07:09,000 Now this is what is the main thing that actually changes in regression. 141 00:07:09,000 --> 00:07:13,000 In loss we will basically use mean absolute error okay. 142 00:07:13,000 --> 00:07:15,000 You can also use mean squared error. 143 00:07:15,000 --> 00:07:18,000 But here we are just going to use mean absolute error. 144 00:07:18,000 --> 00:07:19,000 It is up to you. 145 00:07:19,000 --> 00:07:23,000 And here the metrics that we are specifically going to use is nothing. 146 00:07:23,000 --> 00:07:24,000 But it will be me. 147 00:07:24,000 --> 00:07:24,000 Okay. 148 00:07:25,000 --> 00:07:31,000 Again, you can see the Keras documentation what all loss will specifically be happening over here. 149 00:07:31,000 --> 00:07:34,000 So if I go ahead and open Google right. 150 00:07:34,000 --> 00:07:39,000 And if I just search for Keras loss functions okay. 151 00:07:39,000 --> 00:07:41,000 So if I just go ahead and click on this. 152 00:07:41,000 --> 00:07:48,000 So it shows so many different for, for uh, you know uh classification problems you have binary cross 153 00:07:48,000 --> 00:07:51,000 entropy, binary focus, cross entropy, categorical cross entropy. 154 00:07:52,000 --> 00:07:55,000 Um, then you have binary cross entropy functions, all these things. 155 00:07:55,000 --> 00:07:55,000 Right? 156 00:07:56,000 --> 00:07:58,000 Uh, for regression you have mean squared class. 157 00:07:58,000 --> 00:07:59,000 See this? 158 00:07:59,000 --> 00:08:00,000 This is what I've used. 159 00:08:00,000 --> 00:08:00,000 Right? 160 00:08:00,000 --> 00:08:01,000 Means absolute error. 161 00:08:01,000 --> 00:08:03,000 Right function. 162 00:08:03,000 --> 00:08:05,000 Uh, you can also use mean squared error. 163 00:08:05,000 --> 00:08:05,000 Right. 164 00:08:05,000 --> 00:08:11,000 And already I've spoken about this, uh, even in machine learning, if, you know, machine learning, 165 00:08:11,000 --> 00:08:14,000 I think there all these things will be probably covered up, you know? 166 00:08:14,000 --> 00:08:19,000 So you need to have that basic knowledge with respect to all using this kind of loss function. 167 00:08:19,000 --> 00:08:20,000 Okay. 168 00:08:20,000 --> 00:08:23,000 So now what I will do I will use a matrix as me. 169 00:08:23,000 --> 00:08:25,000 And then finally let's go ahead and display the summary. 170 00:08:25,000 --> 00:08:27,000 So I'll just go ahead and write model dot summary. 171 00:08:27,000 --> 00:08:32,000 And when I display the model summary then here you can see this is my input. 172 00:08:32,000 --> 00:08:33,000 Uh this is my input. 173 00:08:33,000 --> 00:08:38,000 So the total number of weights and bias are nothing but two nine, four, five. 174 00:08:38,000 --> 00:08:40,000 These are the trainable parameters that we have right? 175 00:08:40,000 --> 00:08:43,000 Then I have dense one dense two. 176 00:08:43,000 --> 00:08:45,000 And finally my output is over here okay. 177 00:08:45,000 --> 00:08:50,000 Now let's go ahead and train this particular model again for training this model I will go ahead and 178 00:08:50,000 --> 00:08:53,000 set my logs okay. 179 00:08:53,000 --> 00:09:00,000 So here are some of the logs I'll say this is my regression logs I'm using early stopping and TensorBoard 180 00:09:00,000 --> 00:09:01,000 as you all know. 181 00:09:01,000 --> 00:09:08,000 So we are going to use both of them so that this folder will get uh, it will just try to whatever logs 182 00:09:08,000 --> 00:09:09,000 is basically coming during the training. 183 00:09:09,000 --> 00:09:11,000 It will be coming over here. 184 00:09:11,000 --> 00:09:14,000 Then we are going to set it to early stopping also. 185 00:09:14,000 --> 00:09:17,000 Now in the case of early stopping I will be using validation loss. 186 00:09:17,000 --> 00:09:23,000 Like uh, whenever we are training our neural network you will be getting, um, training loss, uh 187 00:09:23,000 --> 00:09:24,000 sorry training accuracy. 188 00:09:24,000 --> 00:09:25,000 You'll having training loss. 189 00:09:25,000 --> 00:09:28,000 Then you'll also have, uh, validation accuracy and validation loss. 190 00:09:28,000 --> 00:09:30,000 But I really need to monitor the validation loss. 191 00:09:30,000 --> 00:09:32,000 And I'll keep the patient's value to ten. 192 00:09:32,000 --> 00:09:35,000 That basically means I'm going to see it for ten epochs okay. 193 00:09:35,000 --> 00:09:38,000 And finally we will go ahead and train the model. 194 00:09:38,000 --> 00:09:43,000 Now in order to train the model I need to use this callbacks so early stopping and TensorBoard callback 195 00:09:43,000 --> 00:09:47,000 I'm giving my Xtrain Ytrain y epochs will be 100 and all. 196 00:09:47,000 --> 00:09:48,000 Everything is there, right? 197 00:09:48,000 --> 00:09:54,000 So let's go ahead and train this, uh, and let's see how much will be the answer that I will be getting. 198 00:09:54,000 --> 00:09:56,000 So epoch one here, you can see me. 199 00:09:57,000 --> 00:10:02,000 Uh, uh, these all values are there validation loss here also values are coming. 200 00:10:02,000 --> 00:10:03,000 Right. 201 00:10:03,000 --> 00:10:06,000 And these are keep on decreasing unless and until keeps on decreasing. 202 00:10:06,000 --> 00:10:08,000 This epochs will be going on right. 203 00:10:08,000 --> 00:10:13,000 So the validation loss also you'll be able to see that it is decreasing okay. 204 00:10:13,000 --> 00:10:17,000 See validation m e validation loss unless and until is not decreasing. 205 00:10:17,000 --> 00:10:18,000 Right. 206 00:10:18,000 --> 00:10:21,000 If it if it is maintained uh, in a similar manner. 207 00:10:21,000 --> 00:10:22,000 Hardly. 208 00:10:22,000 --> 00:10:23,000 There is not much difference. 209 00:10:23,000 --> 00:10:24,000 Then it will stop. 210 00:10:24,000 --> 00:10:30,000 So here you can see in epoch 47 it has stopped because till here the loss was going down. 211 00:10:30,000 --> 00:10:32,000 But after this it was almost similar. 212 00:10:32,000 --> 00:10:33,000 Right. 213 00:10:33,000 --> 00:10:36,000 So uh, now my model is almost trained. 214 00:10:36,000 --> 00:10:39,000 Uh, again, this is, uh, I'm actually doing it in my local machine. 215 00:10:40,000 --> 00:10:44,000 Again, I have a powerful CPU, so I'll be able to do it very much fast. 216 00:10:44,000 --> 00:10:46,000 Now let's go ahead and evaluate the model. 217 00:10:49,000 --> 00:10:49,000 Okay. 218 00:10:49,000 --> 00:10:57,000 Now, before that, uh, let me just go ahead and, uh, load my, uh, load extension TensorBoard. 219 00:10:57,000 --> 00:10:57,000 Okay. 220 00:10:58,000 --> 00:11:01,000 So I will just go ahead and execute it. 221 00:11:01,000 --> 00:11:09,000 Then I will say, hey, go ahead and take this TensorBoard and my directory will be regression logs. 222 00:11:09,000 --> 00:11:10,000 Slash. 223 00:11:10,000 --> 00:11:12,000 Let's see this particular folder. 224 00:11:12,000 --> 00:11:16,000 So here you have regression logs slash fit okay. 225 00:11:16,000 --> 00:11:19,000 So slash fit will be my. 226 00:11:19,000 --> 00:11:24,000 So once I execute it here will be able to see that we'll be able to load the TensorBoard. 227 00:11:24,000 --> 00:11:29,000 Now in this TensorBoard all the information regarding how the training has happened and all. 228 00:11:29,000 --> 00:11:32,000 So I'm getting some error. 229 00:11:32,000 --> 00:11:33,000 So this should be directory I guess. 230 00:11:33,000 --> 00:11:36,000 Let's see over here okay. 231 00:11:36,000 --> 00:11:38,000 Again you do not need to buy at it. 232 00:11:38,000 --> 00:11:41,000 You can see the documentation and you should be able to see this okay. 233 00:11:41,000 --> 00:11:44,000 You may be thinking Chris you don't know all these things. 234 00:11:44,000 --> 00:11:47,000 Uh I'm not a I will definitely not buy at it. 235 00:11:47,000 --> 00:11:49,000 So you should also not do that okay. 236 00:11:51,000 --> 00:11:54,000 So this should be the log directory. 237 00:11:54,000 --> 00:11:57,000 And here should be my folder okay. 238 00:11:57,000 --> 00:11:59,000 Now I think it should get executed. 239 00:11:59,000 --> 00:12:00,000 Now we are launching the TensorBoard. 240 00:12:00,000 --> 00:12:02,000 Now this is good enough. 241 00:12:02,000 --> 00:12:06,000 It will take all the events from inside this and it will display it to you okay. 242 00:12:06,000 --> 00:12:12,000 And let's see it is taking some time to load it quickly. 243 00:12:12,000 --> 00:12:13,000 Let it get loaded. 244 00:12:14,000 --> 00:12:20,000 I think for the first time I have to execute two times, but it does not happen in Google Colab. 245 00:12:20,000 --> 00:12:23,000 In Google Colab, in the first instance, only you will be able to run it. 246 00:12:23,000 --> 00:12:26,000 Okay, so this is really, really good. 247 00:12:26,000 --> 00:12:31,000 So here I have got my entire information which is amazing in the TensorBoard. 248 00:12:31,000 --> 00:12:37,000 Now see see see this entire things right now if I go ahead and see with respect to regression epoch 249 00:12:37,000 --> 00:12:38,000 versus loss. 250 00:12:38,000 --> 00:12:40,000 Wow this graph looks amazing. 251 00:12:40,000 --> 00:12:42,000 See it is going down. 252 00:12:42,000 --> 00:12:43,000 Down completely. 253 00:12:43,000 --> 00:12:43,000 Right? 254 00:12:43,000 --> 00:12:46,000 So loss is coming down. 255 00:12:46,000 --> 00:12:48,000 If you see mean absolute error this is also coming down. 256 00:12:48,000 --> 00:12:50,000 I think this is an amazing graph. 257 00:12:51,000 --> 00:12:52,000 And this looks good. 258 00:12:52,000 --> 00:12:57,000 That basically means your epoch is going down that the model is really performing well. 259 00:12:57,000 --> 00:12:59,000 Then here you have evaluation loss versus iteration. 260 00:12:59,000 --> 00:13:00,000 No iteration by iteration. 261 00:13:00,000 --> 00:13:02,000 How much it is basically getting reduced. 262 00:13:02,000 --> 00:13:05,000 Then you also have me versus iteration. 263 00:13:05,000 --> 00:13:06,000 This also you can actually check it out, right? 264 00:13:07,000 --> 00:13:10,000 Um, if you just just want to see the validation. 265 00:13:10,000 --> 00:13:12,000 Validation has this epic loss and all. 266 00:13:12,000 --> 00:13:13,000 So sorry epoch loss and all. 267 00:13:13,000 --> 00:13:15,000 So here you have your Ma. 268 00:13:16,000 --> 00:13:17,000 You can probably zoom in and zoom out. 269 00:13:17,000 --> 00:13:22,000 And based on this you can check you can also do something like this and check it out. 270 00:13:22,000 --> 00:13:23,000 Right. 271 00:13:23,000 --> 00:13:26,000 So all this information from the logs you are able to see it okay. 272 00:13:26,000 --> 00:13:28,000 Now let's do one thing. 273 00:13:29,000 --> 00:13:36,000 So if you see me and all, we will also try to evaluate this model for our test data. 274 00:13:36,000 --> 00:13:40,000 So let's go ahead and evaluate model on the test data. 275 00:13:41,000 --> 00:13:42,000 We're just going to check the accuracy how good it is. 276 00:13:42,000 --> 00:13:48,000 So I will just go ahead and compute my test underscore loss and test underscore m. 277 00:13:49,000 --> 00:13:59,000 Uh sorry me mean absolute error I'm just going to write model dot evaluate x underscore test y underscore 278 00:13:59,000 --> 00:14:00,000 test perfect. 279 00:14:00,000 --> 00:14:03,000 So now I'll just go ahead and print. 280 00:14:03,000 --> 00:14:04,000 And I'll use a f string. 281 00:14:04,000 --> 00:14:06,000 And I'll say hey this is my test. 282 00:14:06,000 --> 00:14:13,000 M a e which is nothing but it will be test underscore ma. 283 00:14:14,000 --> 00:14:18,000 So here you can see five zero, one six and more near to zero. 284 00:14:18,000 --> 00:14:23,000 If it is going on, we can probably say that our model is pretty trained in an amazing way. 285 00:14:23,000 --> 00:14:24,000 Okay. 286 00:14:24,000 --> 00:14:25,000 And this is done. 287 00:14:25,000 --> 00:14:33,000 And what you can also do is that, uh, um, along with this, I can also go ahead and save my regression.py 288 00:14:33,000 --> 00:14:34,000 file. 289 00:14:34,000 --> 00:14:39,000 So I will say, hey, go and save this as my regression model. 290 00:14:40,000 --> 00:14:44,000 Regression model dot h5 file. 291 00:14:44,000 --> 00:14:49,000 Okay, so once I do this, this entire model will get saved. 292 00:14:49,000 --> 00:14:52,000 And I have my regression model dot h5 file. 293 00:14:52,000 --> 00:14:52,000 Right. 294 00:14:53,000 --> 00:14:55,000 So almost we have done each and everything right. 295 00:14:55,000 --> 00:14:59,000 Now if you want to deploy this you have all the H5 files. 296 00:14:59,000 --> 00:15:01,000 You have all the training that is been done. 297 00:15:01,000 --> 00:15:02,000 Now you can also go ahead and deploy it. 298 00:15:02,000 --> 00:15:07,000 So for deployment purpose, what I am doing is that I will give you the entire code again with the help 299 00:15:07,000 --> 00:15:09,000 of Streamlit, you can deploy it. 300 00:15:09,000 --> 00:15:11,000 Let me just go ahead and create a file over here. 301 00:15:11,000 --> 00:15:17,000 Let me say streamlit_regression.py okay. 302 00:15:17,000 --> 00:15:21,000 So I'll give you the code over here so that you can go ahead and do the deployment. 303 00:15:21,000 --> 00:15:22,000 And it's very simple. 304 00:15:23,000 --> 00:15:24,000 It's almost like this only. 305 00:15:24,000 --> 00:15:27,000 So here you can see all the things and estimated salary. 306 00:15:27,000 --> 00:15:29,000 You can basically go ahead and compute it okay. 307 00:15:30,000 --> 00:15:32,000 See we are loading it. 308 00:15:32,000 --> 00:15:34,000 We are creating some user input box. 309 00:15:34,000 --> 00:15:36,000 We are converting this into a data frame. 310 00:15:37,000 --> 00:15:38,000 We are doing the encoding. 311 00:15:38,000 --> 00:15:40,000 Then we are predicting the salary okay. 312 00:15:40,000 --> 00:15:42,000 You can do this entire thing okay. 313 00:15:44,000 --> 00:15:46,000 Oh let me see one more model huh. 314 00:15:46,000 --> 00:15:48,000 Regression underscore model dot h5 file okay. 315 00:15:48,000 --> 00:15:49,000 Perfect. 316 00:15:49,000 --> 00:15:52,000 So I hope you understood this particular video. 317 00:15:52,000 --> 00:15:52,000 Again. 318 00:15:52,000 --> 00:15:59,000 It's more about how much you practice and how easy, uh, it was probably to train the entire. 319 00:15:59,000 --> 00:16:03,000 And that is artificial neural network, uh, for classification, for regression. 320 00:16:03,000 --> 00:16:07,000 Now, one of the most common question that has been asked is how do you decide that? 321 00:16:07,000 --> 00:16:11,000 How many hidden layers are neural, uh, hidden neurons you need to take? 322 00:16:11,000 --> 00:16:14,000 We need to discuss about that also, because that is something very important. 323 00:16:14,000 --> 00:16:19,000 So in this case, it is very, very much difficult to just say that, hey, you probably need to just 324 00:16:19,000 --> 00:16:22,000 use this many number of hidden layers or this many number of hidden neurons. 325 00:16:22,000 --> 00:16:23,000 That is not possible. 326 00:16:23,000 --> 00:16:29,000 So what we do is that we basically apply a lot of hyperparameter tuning, like with respect to the number 327 00:16:29,000 --> 00:16:32,000 of hidden layers and uh, with respect to number of hidden neurons. 328 00:16:32,000 --> 00:16:36,000 And that is what I will be showing you in the next video, like how you can also come to a conclusion 329 00:16:36,000 --> 00:16:41,000 and I'll show you one type of hyperparameter tuning how, uh, we can specifically do. 330 00:16:41,000 --> 00:16:43,000 So yes, uh, this was it for my side. 331 00:16:43,000 --> 00:16:44,000 I hope you like this particular video. 332 00:16:44,000 --> 00:16:45,000 I'll see you all in the next video. 333 00:16:45,000 --> 00:16:45,000 Thank you.