如何将以下原始使用情况数据重塑为“分钟数据框”。这样的操作是否有特殊的熊猫功能，可以将原始数据划分为分钟？
原始使用情况数据示例：
**Video-ID | UsageStart** | **Duration in sec** |
0 | 260581 | 2019-04-25 00:00:00 | 10 |
1 | 316288 | 2019-04-25 00:01:05 | 20 |
2 | 791714 | 2019-04-25 00:01:30 | 10 |
3 | 790503 | 2019-04-25 00:02:30 | 90 |
4 | 646034 | 2019-04-25 00:03:10 | 100 |
所需的输出：分钟格式：
**Minute | StartTime | UsageAmount in sec |**
1 | 2019-04-25 00:00:00 | 10 |
2 | 2019-04-25 00:01:00 | 30 |
3 | 2019-04-25 00:02:00 | 30 |
4 | 2019-04-25 00:03:00 | 110 |
5 | 2019-04-25 00:04:00 | 50 |
说实话，我不知道该怎么做。也许这必须先秒秒完成，而不是重新调整为分钟格式。
感谢您的帮助。
答案 0 :(得分：1)
# convert UsageStart to datetime column
df['UsageStart']= pd.to_datetime(df['UsageStart'])
# reindex and sum
df = df.set_index('UsageStart').resample('1T').sum()
答案 1 :(得分：0)
您好，将日期时间列设置为索引后，您可以使用pandas.DataFrame.resample方法，
像
df["UsageStart"] = pd.to_datetime["UsageStart"]
df = df.set_index("UsageStart")
df = df.resample(freq="1Min").mean()
但我不知道平均数能否为您提供所需的输出结果
答案 2 :(得分：0)
这不是一个纯粹的JAVA_HOME environment variable is set to /usr/java/jdk1.8.0_211-amd64
CARBON_HOME environment variable is set to /opt/wso2am-analytics
RUNTIME_HOME environment variable is set to /opt/wso2am-analytics/wso2/dashboard
[2019-04-26 16:16:20,189] INFORMAÇÕES {org.wso2.carbon.launcher.extensions.OSGiLibBundleDeployerUtils updateOSGiLib} - Successfully updated the OSGi bundle information of Carbon Runtime: dashboard
osgi> [2019-04-26 16:16:22,006] INFO {org.wso2.msf4j.internal.websocket.EndpointsRegistryImpl} - Endpoint Registered : /websocket-provider/{topic}
[2019-04-26 16:16:22,559] INFO {org.wso2.msf4j.internal.websocket.WebSocketServerSC} - All required capabilities are available of WebSocket service component is available.
[2019-04-26 16:16:22,571] INFO {org.wso2.carbon.metrics.core.config.model.JmxReporterConfig} - Creating JMX reporter for Metrics with domain 'org.wso2.carbon.metrics'
[2019-04-26 16:16:22,599] INFO {org.wso2.msf4j.internal.MicroservicesServerSC} - All microservices are available
[2019-04-26 16:16:22,599] INFO {org.wso2.carbon.metrics.core.reporter.impl.AbstractReporter} - Started JMX reporter for Metrics
[2019-04-26 16:16:22,669] INFO {org.wso2.msf4j.analytics.metrics.MetricsComponent} - Metrics Component is activated
[2019-04-26 16:16:22,673] INFO {org.wso2.carbon.databridge.agent.internal.DataAgentDS} - Successfully deployed Agent Server
[2019-04-26 16:16:22,742] INFO {org.wso2.transport.http.netty.listener.ServerConnectorBootstrap$HTTPServerConnector} - HTTP(S) Interface starting on host 0.0.0.0 and port 9643
[2019-04-26 16:17:22,535] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component carbon-deployment-service from bundle(org.wso2.carbon.deployment.engine:5.2.0) is in the pending state until Capability org.wso2.carbon.deployment.engine.Deployer from bundle(org.wso2.carbon.uiserver:0.19.5) is available as an OSGi service. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,536] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component carbon-ui-server-startup-listener from bundle(org.wso2.carbon.uiserver:0.19.5) is in the pending state, because of the Capability org.wso2.carbon.uiserver.spi.RestApiProvider from bundle(org.wso2.carbon.business.rules.core:2.0.423). If you've registered this capability as an OSGi service, you need to declare it using the Carbon-Component manifest header. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,536] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component carbon-ui-server-startup-listener from bundle(org.wso2.carbon.uiserver:0.19.5) is in the pending state, because of the Capability org.wso2.carbon.uiserver.spi.RestApiProvider from bundle(org.wso2.carbon.dashboards.api:4.0.38). If you've registered this capability as an OSGi service, you need to declare it using the Carbon-Component manifest header. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,536] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component carbon-ui-server-startup-listener from bundle(org.wso2.carbon.uiserver:0.19.5) is in the pending state until Capability org.wso2.carbon.uiserver.spi.RestApiProvider from bundle(org.wso2.carbon.data.provider:2.0.423) is available as an OSGi service. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,536] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component carbon-ui-server-startup-listener from bundle(org.wso2.carbon.uiserver:0.19.5) is in the pending state until Capability org.wso2.carbon.uiserver.spi.RestApiProvider from bundle(org.wso2.carbon.status.dashboard.core:2.0.423) is available as an OSGi service. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,537] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component sp-idp-service from bundle(org.wso2.carbon.analytics.idp.client:6.0.70) is in the pending state, because of the Capability org.wso2.carbon.analytics.idp.client.core.spi.IdPClientFactory from bundle(org.wso2.carbon.analytics.idp.client:6.0.70). If you've registered this capability as an OSGi service, you need to declare it using the Carbon-Component manifest header. Refer the Startup Order Resolver documentation for information.
[2019-04-26 16:17:22,537] WARN {org.wso2.carbon.kernel.internal.startupresolver.StartupOrderResolver} - Startup component sp-idp-service from bundle(org.wso2.carbon.analytics.idp.client:6.0.70) is in the pending state until Capability org.wso2.carbon.analytics.idp.client.core.spi.IdPClientFactory from bundle(org.wso2.carbon.analytics.idp.client:6.0.70) is available as an OSGi service. Refer the Startup Order Resolver documentation for information.
解决方案，我很确定有很多棘手的oneliner方法可以做到这一点，但是我仍然是熊猫的基本用户。
我使用一种递归函数，该函数通过将给定的pandas的秒数添加到连续的分钟中来消耗该duration，并从d开始将其存储为字典startime的键：
def cumsec(startime, duration, d):
if duration == 0:
return d
to_minute = (60 - startime.second)%60 if (60 - startime.second)%60 else 60
to_add = to_minute if duration - to_minute >= 0 else duration
d[startime.replace(second=0)] += to_add
startime = (startime + dt.timedelta(minutes=1)).replace(second=0)
return cumsec(startime, duration - to_add, d)
然后只需将此功能应用于每一行：
from collections import defaultdict
import datetime as dt
import pandas as pd
# small df arrangements
df.columns = ["VideoId", "UsageStart", "Duration"]
df["UsageStart"] = pd.to_datetime(df["UsageStart"])
d = defaultdict(int)
for r in df.itertuples():
cumsec(r.UsageStart, r.Duration, d)
为了添加可能的空分钟，您可以执行以下操作，但是我敢肯定pandas中有一个特定的方法可以做到这一点（如果不需要这种行为，可以跳过此部分）：< / p>
first = min(d.keys())
last = max(d.keys())
d = {
first + dt.timedelta(minutes=i): d.get(first + dt.timedelta(minutes=i), 0)
for i in range(int((last - first).total_seconds()//60) + 1)
}
最后创建一个新的DataFrame：
cumdf = pd.DataFrame({"StartTime": list(d.keys()), "UsageAmount": list(d.values())})
cumdf = cumdf.sort_values("StartTime").reset_index(drop=True)
cumdf["Minute"] = range(1, len(d) + 1)
print(cumdf)
因此，如果您输入的是：
Video-ID | UsageStart | Duration
459224 | 2019-04-24 23:59:59 | 2
260581 | 2019-04-25 00:00:00 | 10
316288 | 2019-04-25 00:01:05 | 20
791714 | 2019-04-25 00:01:30 | 10
790503 | 2019-04-25 00:02:30 | 90
646034 | 2019-04-25 00:03:10 | 100
934784 | 2019-04-25 00:09:10 | 40
输出为：
StartTime UsageAmount Minute
0 2019-04-24 23:59:00 1 1
1 2019-04-25 00:00:00 11 2
2 2019-04-25 00:01:00 30 3
3 2019-04-25 00:02:00 30 4
4 2019-04-25 00:03:00 110 5
5 2019-04-25 00:04:00 50 6
6 2019-04-25 00:05:00 0 7
7 2019-04-25 00:06:00 0 8
8 2019-04-25 00:07:00 0 9
9 2019-04-25 00:08:00 0 10
10 2019-04-25 00:09:00 40 11